US6768919B2  Method and system for evaluating cardiac ischemia with heart rate feedback  Google Patents
Method and system for evaluating cardiac ischemia with heart rate feedback Download PDFInfo
 Publication number
 US6768919B2 US6768919B2 US10/036,004 US3600401A US6768919B2 US 6768919 B2 US6768919 B2 US 6768919B2 US 3600401 A US3600401 A US 3600401A US 6768919 B2 US6768919 B2 US 6768919B2
 Authority
 US
 United States
 Prior art keywords
 rr
 subject
 qt
 heart rate
 interval
 Prior art date
 Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
 Expired  Fee Related
Links
Images
Classifications

 A—HUMAN NECESSITIES
 A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
 A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
 A61B5/00—Detecting, measuring or recording for diagnostic purposes; Identification of persons
 A61B5/04—Measuring bioelectric signals of the body or parts thereof
 A61B5/0402—Electrocardiography, i.e. ECG
 A61B5/0452—Detecting specific parameters of the electrocardiograph cycle

 A—HUMAN NECESSITIES
 A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
 A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
 A61B5/00—Detecting, measuring or recording for diagnostic purposes; Identification of persons
 A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heartrate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
 A61B5/024—Detecting, measuring or recording pulse rate or heart rate
 A61B5/02405—Determining heart rate variability

 A—HUMAN NECESSITIES
 A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
 A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
 A61B5/00—Detecting, measuring or recording for diagnostic purposes; Identification of persons
 A61B5/04—Measuring bioelectric signals of the body or parts thereof
 A61B5/0402—Electrocardiography, i.e. ECG
 A61B5/0452—Detecting specific parameters of the electrocardiograph cycle
 A61B5/0456—Detecting R peaks, e.g. for synchronising diagnostic apparatus

 A—HUMAN NECESSITIES
 A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
 A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
 A61B5/00—Detecting, measuring or recording for diagnostic purposes; Identification of persons
 A61B5/04—Measuring bioelectric signals of the body or parts thereof
 A61B5/0402—Electrocardiography, i.e. ECG
 A61B5/0452—Detecting specific parameters of the electrocardiograph cycle
 A61B5/0472—Detecting abnormal QRS complex, e.g. widening

 A—HUMAN NECESSITIES
 A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
 A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
 A61B5/00—Detecting, measuring or recording for diagnostic purposes; Identification of persons
 A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heartrate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
 A61B5/024—Detecting, measuring or recording pulse rate or heart rate
 A61B5/0245—Detecting, measuring or recording pulse rate or heart rate by using sensing means generating electric signals, i.e. ECG signals
Abstract
Description
This application is a continuationinpart of copending application Ser. No. 09/891,910, filed Jun. 26, 2001, which is in turn a continuationinpart of application Ser. No. 09/603,286, filed Jun. 26, 2000, now U.S. Pat. No. 6,361,503 the disclosures of both which are incorporated by reference herein in their entirety.
The present invention relates to noninvasive highresolution diagnostics of cardiac ischemia based on processing of bodysurface electrocardiogram (ECG) data. The invention's quantitative method of assessment of cardiac ischemia may simultaneously indicate both cardiac health itself and cardiovascular system health in general.
Heart attacks and other ischemic events of the heart are among the leading causes of death and disability in the United States. In general, the susceptibility of a particular patient to heart attack or the like can be assessed by examining the heart for evidence of ischemia (insufficient blood flow to the heart tissue itself resulting in an insufficient oxygen supply) during periods of elevated heart activity. Of course, it is highly desirable that the measuring technique be sufficiently benign to be carried out without undue stress to the heart (the condition of which might not yet be known) and without undue discomfort to the patient.
The cardiovascular system responds to changes in physiological stress by adjusting the heart rate, which adjustments can be evaluated by measuring the surface ECG RR intervals. The time intervals between consecutive R waves indicate the intervals between the consecutive heartbeats (RR intervals). This adjustment normally occurs along with corresponding changes in the duration of the ECG QT intervals, which characterize the duration of electrical excitation of cardiac muscle and represent the action potential duration averaged over a certain volume of cardiac muscle (FIG. 1). Generally speaking, an average action potential duration measured as the QT interval at each ECG lead may be considered as an indicator of cardiac systolic activity varying in time.
Recent advances in computer technology have led to improvements in automatic analyzing of heart rate and QT interval variability. It is well known now that the QT interval's variability (dispersion) observations performed separately or in combination with heart rate (or RRinterval) variability analysis provides an effective tool for the assessment of individual susceptibility to cardiac arrhythmias (B. Surawicz, J. Cardiovasc. Electrophysiol, 1996, 7, 777784). Applications of different types of QT and some other interval variability to susceptibility to cardiac arrhythmias are described in U.S. Patents by Chamoun U.S. Pat. No. 5,020,540, 1991; Wang U.S. Pat. No. 4,870,974, 1989; Kroll et al. U.S. Pat. No. 5,117,834, 1992; Henkin et al. U.S. Pat. No. 5,323,783, 1994; Xue et al. U.S. Pat. No. 5,792,065, 1998; Lander U.S. Pat. No. 5,827,195, 1998; Lander et al. U.S. Pat. No. 5,891,047, 1999; Hojum et al. U.S. Pat. No. 5,951,484, 1999).
It was recently found that cardiac electrical instability can be also predicted by linking the QT—dispersion observations with the ECG Twave alternation analysis (Verrier et al., U.S. Pat. Nos. 5,560,370; 5,842,997; 5,921,940). This approach is somewhat useful in identifying and managing individuals at risk for sudden cardiac death. The authors report that QT interval dispersion is linked with risk for arrhythmias in patients with long QT syndrome. However, QT interval dispersion alone, without simultaneous measurement of Twave alternation, is said to be a less accurate predictor of cardiac electrical instability (U.S. Pat. No. 5,560,370 at column 6, lines 415).
Another application of the QT interval dispersion analysis for prediction of sudden cardiac death is developed by J. Sarma (U.S. Pat. No. 5,419,338). He describes a method of an autonomic nervous system testing that is designed to evaluate the imbalances between both parasympathetic and sympathetic controls on the heart and, thus, to indicate a predisposition for sudden cardiac death.
The same author suggested that an autonomic nervous system testing procedure might be designed on the basis of the QT hysteresis (J. Sarma et al., PACE 10, 485491 (1988)). Hysteresis between exercise and recovery was observed, and was attributed to sympathoadrenal activity in the early postexercise period. Such an activity was revealed in the course of QT interval adaptation to changes in the RR interval during exercise with rapid variation of the load.
The influence of sympathoadrenal activity and the sharp dependence of this hysteresis on the time course of abrupt QT interval adaptation to rapid changes in the RR interval dynamics radically overshadows the method's susceptibility to the real ischemiclike changes of cardiac muscle electrical parameters and cardiac electrical conduction. Therefore, this type of hysteresis phenomenon would not be useful in assessing the health of the cardiac muscle itself, or in assessing cardiac ischemia.
A similar sympathoadrenal imbalance type hysteresis phenomenon was observed by A. Krahn et al. (Circulation 96, 15511556 (1997)(see FIG. 2 therein)). The authors state that this type of QT interval hysteresis may be a marker for longQT syndrome. However, longQT syndrome hysteresis is a reflection of a genetic defect of intracardiac ion channels associated with exercise or stressinduced syncope or sudden death. Therefore, similar to the example described above, although due to two different reasons, it also does not involve a measure of cardiac ischemia or cardiac muscle ischemic health.
A conventional noninvasive method of assessing coronary artery diseases associated with cardiac ischemia is based on the observation of morphological changes in a surface electrocardiogram during physiological exercise (stress test). A change of the ECG morphology, such as an inversion of the Twave, is known to be a qualitative indication of ischemia. The dynamics of the ECG STsegments are continuously monitored while the shape and slope, as well as STsegment elevation or depression, measured relative to an average base line, are altering in response to exercise load. A comparison of any of these changes with average values of monitored ST segment data provides an indication of insufficient coronary blood circulation and developing ischemia. Despite a broad clinical acceptance and the availability of computerized Holter monitorlike devices for automatic ST segment data processing, the diagnostic value of this method is limited due to its low sensitivity and low resolution. Since the approach is specifically reliable primarily for ischemic events associated with relatively high coronary artery occlusion, its widespread use often results in false positives, which in turn may lead to unnecessary and more expensive, invasive cardiac catheterization.
Relatively low sensitivity and low resolution, which are fundamental disadvantages of the conventional STsegment depression method, are inherent in such method's being based on measuring an amplitude of a body surface ECG signal, which signal by itself does not accurately reflect changes in an individual cardiac cell's electrical parameters normally changing during an ischemic cardiac event. A body surface ECG signal is a composite determined by action potentials aroused from discharge of hundred of thousands of individual excitable cardiac cells. When electrical activity of excitable cells slightly and locally alters during the development of exerciseinduced local ischemia, its electrical image in the ECG signal on the body surface is significantly overshadowed by the aggregate signal from the rest of the heart. Therefore, regardless of physiological conditions, such as stress or exercise, conventional body surface ECG data processing is characterized by a relatively high threshold (lower sensitivity) of detectable ischemic morphological changes in the ECG signal. An accurate and faultless discrimination of such changes is still a challenging signal processing problem.
Accordingly, an object of the present invention is to provide a noninvasive technique for detecting and measuring cardiac ischemia in a patient.
Another object of the invention is to provide a technique for detecting and measuring cardiac ischemia, which technique is not unduly uncomfortable or stressful for the patient.
Another object of the invention is to provide a technique for detecting and measuring cardiac ischemia, which technique may be implemented with relatively simple equipment.
Still another object of the invention is to provide a technique for detecting and measuring cardiac ischemia, which technique is sensitive to low levels of such ischemia.
The present invention overcomes the deficiencies in the conventional STsegment analysis. Although still based on the processing of a body surface ECG signal, it nevertheless provides a highly sensitive and high resolution method for distinguishing changes in cardiac electrical conduction associated with developing cardiac ischemia. In addition to the significant cardiac ischemic changes detectable by the conventional method, the present invention allows one to determine much smaller ischemiainduced conditions and alterations in cardiac electrical conduction. Thus, unlike a conventional STsegment depression ischemic analysis, the method of the present invention opens up opportunities to detect lowlevel cardiac ischemia (undetectable via the regular STsegment method) and also to resolve and monitor small variations of cardiac ischemia. In particular, individuals who would be considered of the same level of cardiac and cardiovascular health according to a conventional ECG evaluation (an STdepression method), will have different measurements if compared according to the method of the present invention, and the cardiac and cardiovascular health of an individual can be quantitatively evaluated, compared and monitored by repeated applications of the method of the present invention.
The present invention is based in part on the discovery that, under certain physiological conditions, QT and/or RRinterval data sets may be interpreted as representing composite dispersionrestitution curves, which characterize the basic dynamic properties of the medium (in this case, cardiac muscle). Indeed, if rapid interval adaptation facilitated by sympathoadrenal activity occurs much faster than gradual heart rate changes following slow alteration of external physiological conditions, then the interval may be considered primarily as a function of a heart rate and/or a preceding cardiac cycle length and does not substantially depend on timedependent sympathoadrenal transients. In such a case a particular interval data set determines a timeindependent, dispersionlike, quasistationary curve which does not substantially depend on rapid adaptational transients and depends primarily on medium electrical parameters.
Based on this discovery, the present invention provides a highly sensitive and high resolution method of assessing cardiac ischemia. This method allows one to detect comparatively small alterations of cardiac muscle electrical excitation properties that develop during even a moderate ischemic condition. For example, consider a gradual heart rate adjustment in a particular human subject in response to slow (quasistationary), thereandback changes of external physiological conditions. Ideally, when a cardiac muscle is supplied by a sufficient amount of oxygen during both gradually increasing and gradually decreasing heart rate stages, the corresponding, thereandback, quasistationary interval curves which result should be virtually identical. However, if ischemia exists, even if only to a very minor extent, there will be alterations of cardiac muscle repolarization and excitation properties for the human subject with the result that one observes a specific quasistationary hysteresis loop. Unlike nonstationary loops (J. Sarma et al., supra (1987); A. Krahn et al., supra (1997)), the quasistationary hystereses of the present invention do not vary substantially versus the course of sympathoadrenal interval adjustment. The domains and shapes of these loops are not significantly affected by timedependent transients rapidly decaying during a transition from one particular heart rate to another; instead, they depend primarily on ischemiainduced changes of medium parameters. The domain encompassed by such a quasistationary hysteresis loop and its shape represent a new quantitative characteristics that indicate cardiac muscle health itself and the health of the cardiovascular system in general. Moreover, any measure of the shape and/or domain enclosed in the hysteresis loop (a measure of a set as defined in the integral theory) possesses the property that any expansion of the domain results in an increase of the measure. Any such mathematical measure can be taken as the new characteristics of cardiac health mentioned above. An arbitrary monotonic function of such a measure would still represent the same measure in another, transformed scale.
A first aspect of the present invention is a method of assessing cardiac ischemia in a subject to provide a measure of cardiovascular health in that subject. The method comprises the steps of:
(a) collecting a first RRinterval data set (e.g., a first QT and RRinterval data set) from the subject during a stage of gradually increasing heart rate;
(b) collecting a second RRinterval data set (e.g., a second QT and RRinterval data set) from the subject during a stage of gradually decreasing heart rate;
(c) comparing said first interval data set to the second interval data set to determine the difference between the data sets; and
(d) generating from the comparison of step (c) a measure of cardiac ischemia during exercise in said subject, wherein a greater difference between said first and second data sets indicates greater cardiac ischemia and lesser cardiovascular health in said subject.
During the periods of gradually increasing and gradually decreasing heart rate the effect of the sympathetic, parasympathetic, and hormonal control on formation of the hysteresis loop is sufficiently small, minimized or controlled so that the ischemic changes are readily detectable. This maintenance is achieved by effecting a gradual increase and gradual decrease in the heart rate, such as, for example, by controlling the heart rate through pharmacological intervention, by direct electrical stimulation of the heart, or by gradually increasing and gradually decreasing exercise loads.
Accordingly, the foregoing method can be implemented in a variety of different ways. A particular embodiment comprises the steps of:
(a) collecting a first RRinterval data set (e.g., a first QT and RRinterval data set) from said subject during a stage of gradually increasing exercise load and gradually increasing heart rate;
(b) collecting a second RRinterval data set (e.g., a second QT and RRinterval data set) from said subject during a stage of gradually decreasing exercise load and gradually decreasing heart rate;
(c) comparing the interval data set to the second interval data set to determine the difference between said data sets; and
(d) generating from said comparison of step (c) a measure of cardiac ischemia during exercise in said subject, wherein a greater difference between said first and second data sets indicates greater cardiac ischemia and lesser cardiovascular health in said subject.
A second aspect of the present invention is a system for assessing cardiac ischemia in a subject to provide a measure of cardiovascular health in that subject. The system comprises:
(a) an ECG recorder for collecting a first RRinterval data set (e.g., a first QT and RRinterval data set) from the subject during a stage of gradually increasing heart rate and collecting a second RRinterval data set (e.g., a second QT and RRinterval data) set from the subject during a stage of gradually decreasing heart rate;
(b) a computer program running in a computer or other suitable means for comparing said first interval data set to the second interval data set to determine the difference between the data sets; and
(c) a computer program running in a computer or other suitable means for generating from said determination of the difference between the data sets a measure of cardiac ischemia during exercise in said subject, wherein a greater difference between the first and second data sets indicates greater cardiac ischemia and lesser cardiovascular health in the subject.
A further aspect of the present invention is a method of assessing cardiac ischemia in a subject to provide a measure of cardiac or cardiovascular health in that subject, the method comprising the steps, performed on a computer system, of:
(a) providing a first RRinterval data set (e.g., a first QT and RRinterval data set) collected from the subject during a stage of gradually increasing heart rate;
(b) providing a second RRinterval data set (e.g., a second QT and RRinterval data set) collected from the subject during a stage of gradually decreasing heart rate;
(c) comparing the first interval data set to the second interval data set to determine the difference between the data sets; and
(d) generating from the comparison of step (c) a measure of cardiac ischemia during stimulation in the subject, wherein a greater difference between the first and second data sets indicates greater cardiac ischemia and lesser cardiac or cardiovascular health in the subject.
The first and second interval data sets may be collected while minimizing the influence of rapid transients due to autonomic nervous system and hormonal control on the data sets. The first and second interval data sets are collected without an intervening rest stage. The generating step may be carried out by generating curves from each of the data sets, and/or the generating step may be carried out by comparing the shapes of the curves from data sets. In a particular embodiment, the generating step is carried out by determining a measure of the domain between the curves. In another particular embodiment, the generating step is carried out by both comparing the shapes of the curves from data sets and determining a measure of the domain between the curves. The method may include the further step of displaying the curves. In one embodiment the comparing step may be carried out by: (i) filtering the first and second interval data sets; (ii) generating a smoothed hysteresis loop from the filtered first and second interval data sets; and then (iii) determining a measure of the domain inside the smoothed hysteresis loop. In another embodiment, the comparing step may be carried out by: (i) filtering the first and second interval data sets; (ii) generating preliminary minimal values for the first and second interval data sets; (iii) correcting the preliminary minimal values; (iv) generating first and second preliminary smoothed curves from each of the filtered data sets; (v) correcting the preliminary smoothed curves; (vi) fitting the preliminary smoothed curves; (vii) generating a smoothed hysteresis loop from the first and second fitted smoothed curves; and then (viii) determining a measure of the domain inside the hysteresis loop. In still another embodiment, the comparing step is carried out by: (i) filtering the first and second interval data sets by moving average smoothing; (ii) generating a smoothed hysteresis loop from the filtered first and second interval data sets; and then (iii) determining a measure of the domain inside the hysteresis loop.
A further aspect of the present invention is a computer system for assessing cardiac ischemia in a subject to provide a measure of cardiac or cardiovascular health in that subject, the system comprising:
(a) means for providing a first RRinterval data set (e.g., a first QT and RRinterval data set) from the subject during a stage of gradually increasing heart rate;
(b) means for providing a second RRinterval data set (e.g., a second QT and RRinterval data set) from the subject during a stage of gradually decreasing heart rate;
(c) means for comparing the first interval data set to the second interval data set to determine the difference between the data sets; and
(d) means for generating from the comparison of step (c) a measure of cardiac ischemia during stimulation in the subject, wherein a greater difference between the first and second data sets indicates greater cardiac ischemia and lesser cardiac or cardiovascular health in the subject.
A still further aspect of the present invention is a computer program product for assessing cardiac ischemia in a subject to provide a measure of cardiac or cardiovascular health in that subject, the computer program product comprising a computer usable storage medium having computer readable program code means embodied in the medium, the computer readable program code means comprising:
(a) computer readable program code means for comparing a first RRinterval data set (e.g., a first QT and RRinterval data set) to a second RRinterval data set (e.g., a second QT and RRinterval data set) to determine the difference between the data sets; and
(b) computer readable program code means for generating from the comparison of step (c) a measure of cardiac ischemia during stimulation in the subject, wherein a greater difference between the first and second data sets indicates greater cardiac ischemia and lesser cardiac or cardiovascular health in the subject.
In another embodiment of the invention, the step of determining a location of ischemia or maximum ischemia in the heart of the subject may also be included.
In another embodiment of the invention, a rest stage following an abrupt stop in exercise can serve as the period of gradually decreasing heart rate in a patient afflicted with coronary artery disease.
While the present invention is described herein primarily with reference to the use of QT and RRinterval data sets, it will be appreciated that the invention may be implemented in simplified form with the use of RRinterval data sets alone. For use in the claims below, it will be understood that the term “RRinterval data set” is intended to be inclusive of both the embodiments of QT and RRinterval data sets and RRinterval data sets alone, unless expressly subject to the proviso that the data set does not include a QTinterval data set.
The present invention is explained in greater detail in the drawings herein and the specification set forth below.
FIG. 1 is a schematic graphic representation of the action potential in cardiac muscle summed up over its volume and the induced electrocardiogram (ECG) recorded on a human's body surface.
FIG. 2A depicts the equations used in a simplified mathematical model of periodic excitation.
FIG. 2B depicts a periodic excitation wave (action potential, u, and instantaneous threshold, v, generated by computer using a simplified mathematical model, the equations of which are set forth in FIG. 2A.
FIG. 2C depicts a family of four composite dispersionrestitution curves corresponding to four values of the medium excitation threshold.
FIG. 3 is a block diagram of an apparatus for carrying out the present method.
FIG. 4A is a block diagram of the processing steps for data acquisition and analysis of the present invention.
FIG. 4B is an alternative block diagram of the processing steps for data acquisition and analysis of the present invention.
FIG. 5 illustrates experimental QTinterval versus RRinterval hysteresis loops for two healthy male (23 year old, thick line and 47 year old, thin line) subjects plotted on the composite dispersionrestitution curve plane.
FIG. 6 provides examples of the QTRR interval hysteresis for two male subjects, one with a conventional ECG STsegment depression (thin line) and one with a history of a myocardial infarction 12 years prior to the test (thick line). The generation of the curves is explained in greater detail in the specification below.
FIG. 7 illustrates sensitivity of the present invention and shows two QTRR interval hysteresis loops for a male subject, the first one (thick lines) corresponds to the initial test during which an STsegment depression on a conventional ECG was observed, and the second one shown by thin lines measured after a period of regular exercise.
FIG. 8 illustrates a comparative cardiac ischemia analysis based on a particular example of a normalized measure of the hysteresis loop area. <CII>=(CII−CII_{min})/(CII_{max}−CII_{min}) (“CII” means “cardiac ischemia index”). O_{I}, X_{I}, and Y_{I }represent human subject data. X_{i }represents data collected from one subject (0.280.35) in a series of tests (day/night testing, run/walk, about two months between tests); exercise peak heart rate ranged from 120 to 135. Y_{i }represents data collected from one subject (0.460.86) in a series of tests (run/walk, six weeks between tests before and after a period of regular exercise stage); exercise peak heart rate ranged from 122 to 146. Black bars indicate a zone (<CII> less than 0.70) in which a conventional ST depression method does not detect cardiac ischemia. The conventional method may detect cardiac ischemia only in a significantly narrower range indicated by high white bars (Y_{2}, Y_{3}, O_{7}:<CII> greater than 0.70).
FIG. 9 illustrates a typical rapid peripheral nervous system and hormonal control adjustment of the QT and RR interval to an abrupt stop in exercise (that is, an abrupt initiation of a rest stage).
FIG. 10 illustrates a typical slow (quasistationary) QT and RR interval adjustment measured during gradually increasing and gradually decreasing cardiac stimulation.
FIG. 11 demonstrates a blockdiagram of the data processing by the method of optimized consolidation of a moving average, exponential and polynomial fitting (Example 10, steps 18).
FIG. 12 demonstrates results of the processing throughout steps 1 to 8 of Example 10. Upper panels show QT and RR data sets processed from steps 1 to 3 (from left to right respectively), and the QT/RR hysteresis loop after step 1. The exponential fitting curves (step 3) are shown in gray in the first two panels. Low panels show the same smooth dependencies after processing from step 4 to a final step 8. Here the CII (see right low panel) is equal to a ratio S/{((T_{RR} ^{−}(t_{end}−t_{min} ^{RR})−T_{RR} ^{−}(0))((T_{QT} ^{+}(t_{start}−t_{min} ^{QT})−T_{QT} ^{+}(0))}(see example 10, section 7).
FIG. 13 demonstrates a blockdiagram of the data processing by the method of a sequential moving average (Example 11, steps 13)
FIG. 14 demonstrates results of the processing throughout steps 1 to 2 of Example 11. Upper panels show processed QT and RR data sets, and the QT/RR hysteresis loop after step 1 (from left to right respectively). Low panels show the same smooth dependencies after the second moving average processing and a final step 3.
FIG. 15 shows a general data flow chart for major steps in optimized nonlinear transformation method. The lefthand side and the righthand side boxes describe similar processing stages for the RR and QTintervals, respectively.
FIG. 16 shows a detailed data flow chart for one data subset, {t^{k},T^{k} _{RR}} or {t^{k},T^{k} _{QT}} during stages (1 a/b) through (3 a/b) in FIG. 15. The preliminary stage, boxes 1 through 7, uses a combination of traditional data processing methods and includes: moving averaging (1), determination of a minimum region (2), fitting a quadratic parabola to the data in this region (3), checking consistency of the result (4), finding the minimum and centering data at the minimum (5) and (6), conditionally sorting the data (7). Stages (8) through (11) are based on the dualnonlinear transformation method for the nonlinear regression.
FIG. 17 displays the nonlinear transformation of a filtered RRinterval data set. Panel A shows the data on the original (t,y)plane, the minimum is marked with an asterisk inside a circle. Panels B and C show the transformed sets on the (t,u)plane, for j=1 and for j=2, respectively. The image of the minimum is also marked with an encircled asterisk. Note that the transformed data sets concentrate around a monotonously growing (average) curve with a clearly linear portion in the middle.
FIG. 18 is similar to FIG. 17 but for a QTinterval data set.
FIG. 19 displays an appropriately scaled representation for the family of functions ξ(α,β,τ) for fifteen values of parameter β varying with the step Δβ=0.1 from β=−0.9 (lower curve) through β=0 (medium, bold curve), to β=0.5 (upper curve). The function ξ(α,β,τ) is continuous in all three variables and as a function of τ has a unit slope at τ=0, ξ′(α,β,0)=1.
FIG. 20 shows an example of full processing of the RR and QT data sets for one patient. Panels A and C represent RR and QT data sets and their fit. Panel B shows the corresponding ascending and descending curves and closing line on the (T_{RR},T_{QT})plane, on which the area of such a hysteresis loop has the dimension of timesquared. Panel D shows a hysteresis loop on the (ƒ_{RR},T_{QT})plane, where ƒ_{RR}=1/T_{RR }is the heart rate, on which the loop area is dimensionless. The total errorfor Panel A is 2.2% and for Panel C is 0.8%.
FIG. 21 schematically illustrates an apparatus of the present invention incorporating heart rate feedback to control the stages of gradually increasing and gradually decreasing heart rate.
The present invention is explained in greater detail below. This description is not intended to be a detailed catalog of all the different manners in which particular elements of the invention can be implemented, and numerous variations will be apparent to those skilled in the art based upon the instant disclosure.
As will be appreciated by one of skill in the art, certain aspects of the present invention may be embodied as a method, data processing system, or computer program product. Accordingly, certain aspects of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, certain aspects of the present invention may take the form of a computer program product on a computerusable storage medium having computer readable program code means embodied in the medium. Any suitable computer readable medium may be utilized including, but not limited to, hard disks, CDROMs, optical storage devices, and magnetic storage devices.
Certain aspects of the present invention are described below with reference to flowchart illustrations of methods, apparatus (systems), and computer program products. It will be understood that each block of the flowchart illustrations, and combinations of blocks in the flowchart illustrations, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart block or blocks.
Computer program instructions may also be stored in a computerreadable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computerreadable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart block or blocks.
Computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.
1. Definitions
“Cardiac ischemia” refers to a lack of or insufficient blood supply to an area of cardiac muscle. Cardiac ischemia usually occurs in the presence of arteriosclerotic occlusion of a single or a group of coronary arteries. Arteriosclerosis is a product of a lipid deposition process resulting in fibrofatty accumulations, or plaques, which grow on the internal walls of coronary arteries. Such an occlusion compromises blood flow through the artery, which reduction then impairs oxygen supply to the surrounding tissues during increased physiological need—for instance, during increased exercise loads. In the later stages of cardiac ischemia (e.g., significant coronary artery occlusion), the blood supply may be insufficient even while the cardiac muscle is at rest. However, in its earlier stages such ischemia is reversible in a manner analogous to how the cardiac muscle is restored to normal function when the oxygen supply to it returns to a normal physiological level. Thus, ischemia that may be detected by the present invention includes episodic, chronic and acute ischemia.
“Exercise” as used herein refers to voluntary skeletal muscle activity of a subject that increases heart rate above that found at a sustained stationary resting state. Examples of exercise include, but are not limited to, cycling, rowing, weightlifting, walking, running, stairstepping, etc., which may be implemented on a stationary device such as a treadmill or in a nonstationary environment.
“Exercise load” or “load level” refers to the relative strenuousness of a particular exercise, with greater loads or load levels for a given exercise producing a greater heart rate in a subject. For example, load may be increased in weightlifting by increasing the amount of weight; load may be increased in walking or running by increasing the speed and/or increasing the slope or incline of the walking or running surface; etc.
“Gradually increasing” and “gradually decreasing” an exercise load refers to exercise in which the subject is caused to perform an exercise under a plurality of different sequentially increasing or sequentially decreasing loads. The number of steps in the sequence can be infinite so the terms gradually increasing and gradually decreasing loads include continuous load increase and decrease, respectively.
“Intervening rest”, when used to refer to a stage following increased cardiac stimulation, refers to a stage of time initiated by a sufficiently abrupt decrease in heart stimulation (e.g., an abrupt decrease in exercise load) so that it evokes a clear sympathoadrenal response. Thus, an intervening rest stage is characterized by a rapid sympathoadrenal adjustment (as further described in Example 8 below), and the inclusion of an intervening rest stage precludes the use of a quasistationary exercise (or stimulation) protocol (as further described in Example 9 below).
“Hysteresis” refers to a lagging of the physiological effect when the external conditions are changed.
“Hysteresis curves” refer to a pair of curves in which one curve reflects the response of a system to a first sequence of conditions, such as gradually increasing heart rate, and the other curve reflects the response of a system to a second sequence of conditions, such as gradually decreasing heart rate. Here both sets of conditions are essentially the same—i.e., consist of the same (or approximately the same) steps—but are passed in different order in the course of time. A “hysteresis loop” refers to a loop formed by the two contiguous curves of the pair.
“Electrocardiogram” or “ECG” refers to a continuous or sequential record (or a set of such records) of a local electrical potential field obtained from one or more locations outside the cardiac muscle. This field is generated by the combined electrical activity (action potential generation) of multiple cardiac cells. The recording electrodes may be either subcutaneously implanted or may be temporarily attached to the surface of the skin of the subject, usually in the thoracic region. An ECG record typically includes the singlelead ECG signal that represents a potential difference between any two of the recording sites including the site with a zero or ground potential.
“Quasistationary conditions” refer to a gradual change in the external conditions and/or the physiological response it causes that occurs much slower than any corresponding adjustment due to sympathetic/parasympathetic and hormonal control. If the representative time of the external conditions variation is denoted by τ_{ext}, and τ_{int }is a representative time of the fastest of the internal, sympathetic/parasympathetic and hormonal control, then “quasistationary conditions” indicates τ_{ext}>>τ_{int }(e.g., τ_{ext }is at least about five times greater than τ_{int}). “An abrupt change” refers to an opposite situation corresponding to a sufficiently fast change in the external conditions as compared with the rate sympathetic/parasympathetic and hormonal control—that is, it requires that τ_{ext}<<τ_{int }(e.g., ext is at least about five times less than τ_{int}). In particular, “an abrupt stop” refers to a fast removal of the exercise load that occurs during time shorter than τ_{int}˜20 or 30 seconds (see FIG. 9 below and comments therein).
“QT and RR data set” refers to a record of the time course of an electrical signal comprising action potentials spreading through cardiac muscle. Any single lead ECG record incorporates a group of three consecutive sharp deflections usually called a QRS complex and generated by the propagation of the action potential's front through the ventricles. In contrast, the electrical recovery of ventricular tissue is seen on the ECG as a relatively small deflection known as the T wave. The time interval between the cardiac cycles (i.e., between the maxima of the consecutive Rwaves) is called an RRinterval, while the action potential duration (i.e., the time between the beginning of a QRS complex and the end of the ensuing Twave) is called a QTinterval. Alternative definitions of these intervals can be equivalently used in the framework of the present invention. For example, an RRinterval can be defined as the time between any two similar points, such as the similar inflection points, on two consecutive Rwaves, or any other way to measure cardiac cycle length. A QTinterval can be defined as the time interval between the peak of the Qwave and the peak of the T wave. It can also be defined as the time interval between the beginning (or the center) of the Qwave and the end of the ensuing Twave defined as the point on the time axis (the base line) at which it intersects with the linear extrapolation of the Twave's falling branch and started from its inflection point, or any other way to measure action potential duration. An ordered set of such interval durations simultaneously with the time instants of their beginnings or ends which are accumulated on a beat to beat basis or on any given beat sampling rate basis form a corresponding QT and RRinterval data set. Thus, a QT and RRinterval data set will contain two QTinterval related sequences {T_{QT,1}, T_{QT,2}, . . . , T_{QT,n},} and {t_{1}, t_{2}, . . . , t_{n}}, and will also contain two RRinterval related sequences {T_{RR,1}, T_{RR,2}, . . . , T_{RR,n},} and {t_{1}, t_{2}, . . . , t_{n}} (the sequence {t_{1}, t_{2}, . . . , t_{n}} may or may not exactly coincide with the similar sequence in the QT data set).
In the following definitions, C[a,b] shall denote a set of continuous functions ƒ(t) on a segment [a,b]. {t_{i}}, i=1,2, . . . , N, denotes a set of points from [a,b], i.e. {t_{i}}={t_{i}: a≦t_{i}≦b, i=1,2, . . . , N} and {ƒ(t_{i})}, where ƒεC[a,b], denotes a set of values of the function ƒ fat the points {t_{i}}. In matrix operations the quantities τ={t_{i}}, y={ƒ(t_{i})}, are treated as column vectors. E_{N }shall denote a Ndimensional metric space with the metric R_{N}(x,y), x,yεE_{N}. (R_{N}(x,y) is said to be a distance between points x and y.) A (total) variation
is defined for any absolutely continuous function F from C[a,b] as the integral (a Stieltjes integral)
For a function F monotonic on segment [a,b] its variation is simply F(a)−F(b). If a function F(t) has alternating maxima and minima, then the total variation of F is the sum of its variations on the intervals of monotonicity. For example, if the points of minima and maxima are x_{1}=a, x_{2}, x_{3}, . . . , x_{k}=b then
Fitting (Best Fitting)
Let C[a, b] be a subset of C[a,b]. A continuous function ƒ(t), ƒε{tilde over (C)}[a,b] is called the (best) fit (or the best fitting) function of class {tilde over (C)}[a, b] with respect to metric R_{N }to a data set {x_{i},t_{i}} (i=1,2, . . . , N} if
The minimum value of R_{N }is then called the error of the fit. The functions ƒ(t) from {tilde over (C)}[a,b] will be called trial functions.
In most cases E_{N }is implied to be an Euclidean space with an Euclidean metric. The error R_{N }then becomes the familiar meanrootsquare error. The fit is performed on a subset {tilde over (C)}[a,b] since it usually implies a specific parametrization of the trial functions and/or such constrains as the requirements that the trial functions pass through a given point and/or have a given value of the slope at a given point.
A Smoother Function (Comparison of Smoothness)
Let ƒ(t) and g(t) be functions from C[a,b] that have absolutely continuous derivatives on this segment. The function ƒ(t) is smoother than the function g(t) if
where the prime denotes a time derivative, and a strict inequality holds in at least one of relations (D.4) and (D.5).
A Smoother Set
A set {x_{i},t_{i}} (i=1,2, . . . , N} is smoother than the set {x′_{j},t′_{j}} (j=1,2, . . . , N′} if the former can be fit with a smoother function ƒ(t) of the same class within the same or smaller error than the latter.
Smoothing of a Data Set
A (linear) transformation of a data set (x,t)≡{x_{i},t_{i}} (i=1,2, . . . , N_{0}} into another set (y,τ)≡{y_{i},τ_{j}} (j=1,2, . . . , N_{l},} of the form
y=A·x, τ=B·t, (D.6)
where A and B are N_{1}×N_{0 }matrices, is called a smoothing if the latter set is smoother than the former. One can refer to {y_{j},τ_{j}} as a smoothed set
A Measure of a Closed Domain
Let Ω be a singly connected domain on the plane (τ,T) with the boundary formed by a simple (i.e., without selfintersections) continuous curve. A measure M of such a domain Ω on the plane (τ,T) is defined as the Riemann integral
where ρ(τ,T) is a nonnegative (weight) function on Ω.
Note that when ρ(τ,T)≡1 the measure M of the domain coincides with its area, A; when ρ(τ,T)≡1/τ^{2}, the measure, M has the meaning of the area, A′ of the domain Ω′ on the transformed plane (ƒ,T), where ƒ=1/τ can be understood as the heart rate since the quantity τ has the meaning of RRinterval. [The domain Ω′ is the image of domain Ω under the mapping (τ,T)→(1/τ,T).]
2. Dispersion/restitution Curves.
FIG. 1 illustrates the correspondence between the temporal phases of the periodic action potential (AP, upper graph, 20) generated inside cardiac muscle and summed up over its entire volume and the electrical signal produced on the body surface and recorded as an electrocardiogram (ECG, lower graph, 21). The figure depicts two regular cardiac cycles. During the upstroke of the action potential the QRScomplex is formed. It consists of three waves, Q, R, and S, which are marked on the lower panel. The recovery stage of the action potential is characterized by its fall off on the AP plot and by the Twave on the ECG plot. One can see that the action potential duration is well represented by the time between Q and T waves and is conventionally defined as the QT interval, measured from the beginning of the Q wave to the end of the following T wave. The time between consecutive Rwaves (RR interval) represents the duration of a cardiac cycle, while its reciprocal value represents the corresponding instantaneous heart rate.
FIG. 2 illustrates major aspects of the process of propagation of a periodic action potential through cardiac tissue and the formation of a corresponding composite dispersionrestitution curve. The tissue can be considered as a continuous medium and the propagation process as a repetition at each medium point of the consecutive phases of excitation and recovery. The former phase is characterized by a fast growth of the local membrane potential (depolarization), and the latter by its return to a negative resting value (repolarization). The excitation phase involves a very fast (˜0.1 ms) decrease in the excitation threshold and the following development of a fast inward sodium current that causes the upstroke of the action potential (˜1 ms). Next, during an intermediate plateau phase (˜200 ms) sodium current is inactivated; calcium and potassium currents are developing while the membrane is temporarily unexcitable (i.e., the threshold is high). During the next recovery phase (˜100 ms), a potassium current repolarizes the membrane so it again becomes excitable (the excitation threshold is lowered).
The complicated description of a multitude of ionic currents involved in the process can be circumvented if one treats the process directly in terms of the local membrane potential, u, and a local excitation threshold, v. Such a mathematical description referred to as the CSC model, was developed by Chernyak, Starobin, & Cohen (Phys. Rev. Lett., 80, pp.56755678, 1998) and is presented as a set of two ReactionDiffusion (RD) equations in panel A. The lefthand side of the first equation describes local accumulation of the electric charge on the membrane, the first term in the righthand side describes Ohmic coupling between neighboring points of the medium, and the term i(u,v) represents the transmembrane current as a function of the membrane potential and the varying excitation threshold (ε is a small constant, the ratio of the slow recovery rate to the fast excitation rate). A periodic solution (a wavetrain) can be found analytically for some particular functions i(u,v) and g(u,v). The wavetrain shown in panel B has been calculated for g(u,v)=ζu+v_{r}−v, where ζ and v_{r }are appropriately chosen constants (v_{r }has the meaning of the initial excitation threshold and is the main determinant of the medium excitability). The function i(u,v) was chosen to consist of two linear pieces, one for the subthreshold region, u<v, and one for suprathreshold region, u>v. That is i(u,v)=λ_{r}u when u≦v, and i(u,v)=λ_{ex}(u−u_{ex}) when u>v, where λ_{r }and λ_{ex }are membrane chord conductances in the resting (u=0) and excited (u=u_{ex}) states, respectively. The resting state u=0 is taken as the origin of the potential scale. We used such units that λ_{ex}=1 and u_{ex}=1. (For details see Chernyak & Starobin, Critical Reviews in Biomed. Eng. 27, 359414 (1999)).
A medium with higher excitability, corresponding to the tissue with better conduction, gives rise to a faster, more robust action potential with a longer Action Potential Duration (APD). This condition also means that a longerlasting excitation propagates faster. Similarly, a wave train with a higher frequency propagates slower since the medium has less time to recover from the preceding excitation and thus has a lower effective excitability. These are quite generic features that are incorporated in the CSC model. In physics, the relation between the wave's speed, c, and its frequency, ƒ or its period, T=1/ƒ, is called a dispersion relation. In the CSC model the dispersion relation can be obtained in an explicit form T=F_{T}(c), where F_{T }is a known function of c and the medium parameters. The CSC model also allows us to find a relation between the propagation speed and the APD, T_{AP}, in the explicit form T_{AP}=F_{AP}(c), which represents the restitution properties of the medium. In the medical literature, the restitution curve is T_{AP }versus diastolic interval T_{DI}, which differently makes a quite similar physical statement. One can consider a pair of dispersion and restitution relations {T=F_{T}(c),T_{AP}=F_{AP}(c)} as a parametric representation of a single curve on the (T,T_{AP})plane as shown in panel C (FIG. 2). Such a curve (relation) shall be referred to as a composite dispersionrestitution curve (relation) and can be directly obtained from an experimental ECG recording by determining the QTRR interval data set and plotting T_{QT }versus T_{RR}. A condition that the experimental {T_{QT},T_{RR}} data set indeed represents the composite dispersionrestitution relation is the requirement that the data are collected under quasistationary conditions. Understanding this fact is a key discovery for the present invention.
3. Testing Methods
The methods of the present invention are primarily intended for the testing of human subjects. Virtually any human subject can be tested by the methods of the present invention, including male, female, juvenile, adolescent, adult, and geriatric subjects. The methods may be carried out as an initial screening test on subjects for which no substantial previous history or record is available, or may be carried out on a repeated basis on the same subject (particularly where a comparative quantitative indicium of an individual's cardiac health over time is desired) to assess the effect or influence of intervening events and/or intervening therapy on that subject between testing sessions.
As noted above, the method of the present invention generally comprises (a) collecting a first QT and RRinterval data set from said subject during a stage of gradually increasing heart rate; (b) collecting a second QT and RRinterval data set from said subject during a stage of gradually decreasing heart rate; (c) comparing said first QT and RRinterval data set to said second QT and RRinterval data set to determine the difference between said data sets; and (d) generating from said comparison of step (c) a measure of cardiac ischemia in the subject. A greater difference between the first and second data sets indicates greater cardiac ischemia and lesser cardiac or cardiovascular health in that subject.
The stages of gradually increasing and gradually decreasing heart rate are carried out in a manner that maintains during both periods essentially or substantially the same stimulation of the heart by the peripheral nervous and hormonal control systems, so that it is the effect of cardiac ischemia rather than that of the external control which is measured by means of the present invention. This methodology can be carried out by a variety of techniques, with the technique of conducting two consecutive stages of gradually increasing and gradually decreasing exercise loads (or average heart rates) being currently preferred.
The stage of gradually increasing exercise load (or increased average heart rate) and the stage of gradually decreasing exercise load (or decreased average heart rate) may be the same in duration or may be different in duration. In general, each stage is at least 3, 5, 8, or 10 minutes or more in duration. Together, the duration of the two stages may be from about 6, 10, 16 or 20 minutes in duration to about 30, 40, or 60 minutes in duration or more. The two stages are preferably carried out sequentially in time—that is, with one stage following after the other substantially immediately, without an intervening rest stage. In the alternative, the two stages may be carried out separately in time, with an intervening “plateau” stage (e.g., of from 1 to 5 minutes) during which cardiac stimulation or exercise load is held substantially constant, before the stage of decreasing load is initiated.
The exercise protocol may include the same or different sets of load steps during the stages of increasing or decreasing heart rates. For example, the peak load in each stage may be the same or different, and the minimum load in each stage may be the same or different. In general, each stage consists of at least two or three different load levels, in ascending or descending order depending upon the stage. Relatively high load levels, which result in relatively high heart rates, can be used but are not essential. An advantage of the present invention is that its sensitivity allows both exercise procedures to be carried out at relatively low load levels that do not unduly increase the pulse rate of the subject. For example, the method may be carried out so that the heart rate of the subject during either the ascending or descending stage (or both) does not exceed about 140, 120, or even 100 beats per minute, depending upon the condition of the subject. Of course, data collected at heart rates above 100, 120, or 140 beats per minute may also be utilized if desired, again depending upon the condition of the subject.
For example, for an athletic or trained subject, for the first or ascending stage, a first load level may be selected to require a power output of 60 to 100 or 150 watts by the subject; an intermediate load level may be selected to require a power output of 100 to 150 or 200 watts by the subject; and a third load level may be selected to require a power output of 200 to 300 or 450 watts or more by the subject. For the second or descending stage, a first load level may be selected to require a power output of 200 to 300 or 450 watts or more by the subject; an intermediate or second load level may be selected to require a power output of 100 to 150 or 200 watts by the subject; and a third load level may be selected to require a power output of 60 to 100 or 150 watts by the subject. Additional load levels may be included before, after, or between all of the foregoing load levels as desired, and adjustment between load levels can be carried out in any suitable manner, including stepwise or continuously.
In a further example, for an average subject or a subject with a history of cardiovascular disease, for the first or ascending stage, a first load level may be selected to require a power output of 40 to 75 or 100 watts by the subject; an intermediate load level may be selected to require a power output of 75 to 100 or 150 watts by the subject; and a third load level may be selected to require a power output of 125 to 200 or 300 watts or more by the subject. For the second or descending stage, a first load level may be selected to require a power output of 125 to 200 or 300 watts or more by the subject; an intermediate or second load level may be selected to require a power output of 75 to 100 or 150 watts by the subject; and a third load level may be selected to require a power output of 40 to 75 or 100 watts by the subject. As before, additional load levels may be included before, after, or between all of the foregoing load levels as desired, and adjustment between load levels can be carried out in any suitable manner, including stepwise or continuously.
The heart rate may be gradually increased and gradually decreased by subjecting the patient to a predetermined schedule of stimulation. For example, the patient may be subjected to a gradually increasing exercise load and gradually decreasing exercise load, or gradually increasing electrical or pharmacological stimulation and gradually decreasing electrical or pharmacological stimulation, according to a predetermined program or schedule. Such a predetermined schedule is without feedback of actual heart rate from the patient. In the alternative, the heart rate of the patient may be gradually increased and gradually decreased in response to actual heart rate data collected from concurrent monitoring of said patient. Such a system is a feedback system. For example, the heart rate of the patient may be monitored during the test and the exercise load (speed and/or incline, in the case of a treadmill) can be adjusted so that the heart rate varies in a prescribed way during both stages of the test. The monitoring and control of the load can be accomplished by a computer or other control system using a simple control program and an output panel connected to the control system and to the exercise device that generates an analog signal to the exercise device. One advantage of such a feedback system is that (if desired) the control system can insure that the heart rate increases substantially linearly during the first stage and decreases substantially linearly during the second stage. Methods and apparatus incorporating such feedback are explained in greater detail in Example 14 below and FIG. 21 herein.
The generating step (d) may be carried out by any suitable means, such as by generating curves from the data sets (with or without actually displaying the curves), and then (i) directly or indirectly evaluating a measure (e.g., as defined in the integral theory) of the domain (e.g., area) between the hysteresis curves, a greater measure indicating greater cardiac ischemia in said subject, (ii) directly or indirectly comparing the shapes (e.g., slopes or derivatives thereof) of the curves, with a greater difference in shape indicating greater cardiac ischemia in the subject; or (iii) combinations of (i) and (ii). Specific examples are given in Example 4 below.
The method of the invention may further comprise the steps of (e) comparing the measure of cardiac ischemia during exercise to at least one reference value (e.g., a mean, median or mode for the quantitative indicia from a population or subpopulation of individuals) and then (ƒ) generating from the comparison of step (e) at least one quantitative indicium of cardiovascular health for said subject. Any such quantitative indicium may be generated on a onetime basis (e.g., for assessing the likelihood that the subject is at risk to experience a future ischemiarelated cardiac incident such as myocardial infarction or ventricular tachycardia), or may be generated to monitor the progress of the subject over time, either in response to a particular prescribed cardiovascular therapy, or simply as an ongoing monitoring of the physical condition of the subject for improvement or decline (again, specific examples are given in Example 4 below). In such a case, steps (a) through (ƒ) above are repeated on at least one separate occasion to assess the efficacy of the cardiovascular therapy or the progress of the subject. A decrease in the difference between said data sets from before said therapy to after said therapy, or over time, indicates an improvement in cardiac health in said subject from said cardiovascular therapy. Any suitable cardiovascular therapy can be administered, including but not limited to, aerobic exercise, muscle strength building, change in diet, nutritional supplement, weight loss, smoking cessation, stress reduction, pharmaceutical treatment (including gene therapy), surgical treatment (including both open heart and closed heart procedures such as bypass, balloon angioplasty, catheter ablation, etc.) and combinations thereof.
The therapy or therapeutic intervention may be one that is approved or one that is experimental. In the latter case, the present invention may be implemented in the context of a clinical trial of the experimental therapy, with testing being carried out before and after therapy (and/or during therapy) as an aid in determining the efficacy of the proposed therapy.
4. Testing Apparatus
FIG. 3 provides an example of the apparatus for data acquisition, processing and analysis by the present invention. Electrocardiograms are recorded by an ECG recorder 30, via electrical leads placed on a subject's body. The ECG recorder may be, for example, a standard multilead Holter recorder or any other appropriate recorder. The analog/digital converter 31 digitizes the signals recorded by the ECG recorder and transfers them to a personal computer 32, or other computer or central processing unit, through a standard external input/output port. The digitized ECG data can then be processed by standard computerbased waveform analyzer software. Composite dispersionrestitution curves and a cardiac or cardiovascular health indicium or other quantitative measure of the presence, absence or degree of cardiac ischemia can then be calculated automatically in the computer through a program (e.g., Basic, Fortran, C++, etc.) implemented therein as software, hardware, or both hardware and software.
FIG. 4A and FIG. 4B illustrate the major steps of digitized data processing in order to generate an analysis of a QTRR data set collected from a subject during thereandback quasistationary changes in physiological conditions. The first four steps in FIG. 4A and FIG. 4B are substantially the same. The digitized data collected from a multilead recorder are stored in a computer memory for each lead as a data array 40 a, 40 b. The size of each data array is determined by the durations of the ascending and descending heart rate stages and a sampling rate used by the waveform analyzer, which processes an incoming digitized ECG signal. The waveform analyzer software first detects major characteristic waves (Q, R, S and T waves) of the ECG signal in each particular lead 41 a, 41 b. Then in each ECG lead it determines the time intervals between consecutive R waves and the beginning of Q and the end of T waves 42 a, 42 b. Using these reference points it calculates heart rate and RR and QTintervals. Then, the application part of the software sorts the intervals for the ascending and descending heart rate stages 43 a, 43 b. The next two steps can be made in one of the two alternative ways shown in FIGS. 4A and 4B, respectively. The fifth step as shown in FIG. 4A consists of displaying by the application part of software QTintervals versus RRintervals 44 a, separately for the ascending and descending heart rate stages effected by thereandback gradual changes in physiological conditions such as exercise, pharmacological/electrical stimulation, etc. The same part of the software performs the next step 45 a, which is smoothing, filtering or data fitting, using exponential or any other suitable functions, in order to obtain a sufficiently smooth curve T_{QT}=F(T_{RR}) for each stage. An alternative for the last two steps shown in FIG. 4B requires that the application part of the software first averages, and/or filters and/or fits, using exponential or any other suitable functions, the QT intervals as functions of time for both stages and similarly processes the RRinterval data set to produce two sufficiently smooth curves T_{QT}=F_{QT}(t) and T_{RR}=F_{RR}(t), each including the ascending and descending heart rate branches 44 b. At the next step 45 b the application part of the software uses this parametric representation to eliminate time and generate and plot a sufficiently smooth hysteresis loop T_{QT}=F(T_{RR}). The following steps shown in FIGS. 4A and FIG. 4B are again substantially the same. The next step 46 a, 46 b performed by the application part of the software can be graphically presented as closing the two branch hysteresis loop with an appropriate interconnecting or partially connecting line, such as a vertical straight line or a line connecting the initial and final points, in order to produce a closed hysteresis loop on the (T_{QT},T_{RR})plane. At the next step 47 a, 47 b the application software evaluates for each ECG lead an appropriate measure of the domain inside the closed hysteresis loop. A measure, as defined in mathematical integral theory, is a generalization of the concept of an area and may include appropriate weight functions increasing or decreasing the contribution of different portions of the domain into said measure. The final step 48 a, 48 b of the data processing for each ECG lead is that the application software calculates indexes by appropriately renormalizing the said measure or any monotonous functions of said measure. The measure itself along with the indexes may reflect both the severity of the exerciseinduced ischemia, as well as a predisposition to local ischemia that can be reflected in some particularities of the shape of the measured composite dispersionrestitution curves. The results of all abovementioned signal processing steps may be used to quantitatively assess cardiac ischemia and, as a simultaneous option, cardiovascular system health of a particular individual under test.
Instead of using the (T_{QT},T_{RR})plane a similar data processing procedure can equivalently be performed on any plane obtained by a nondegenerate transformation of the (T_{QT},T_{RR})plane, such as (T_{QT},ƒ_{RR}) where ƒ_{RR}=1/T_{RR }is the heart rate or the like. Such a transformation can be partly or fully incorporated in the appropriate definition of the said measure.
5. Ischemia Location
Data sets, including both QT and RRdata sets and RRdata sets alone, can be collected from a single or plurality of different electrodes and leads to a subject. Typically the electrodes forming any number of lead configurations are cutaneous electrodes. For example, the apparatus may be provided with 3, 5, 7, and 12 leads options for collecting the data sets. The most detailed 12lead option includes 6precordial electrodes, with three electrodes in each of the left and right heart respectively, and 6 limb electrodes with three electrodes in each “true” and “aV” configuration.
Quasistationary hysteresis loops can be assessed with the data provided from each of these plurality of electrodes separately, or with data provided from specific combinations of electrodes in groups such as right/left precordial, true, and aV configurations. The location of cardiac ischemia, or greatest cardiac ischemia, can then be determined with data from different electrodes to allow one to find greater or maximum indications of ischemia from data corresponding to a particular electrode or electrode set. The greater or maximum indication of ischemia in data corresponding to a lead or lead set indicates a region of the most developed ischemia in the heart of the subject. Such a region may be localized, for instyance, in the left and/or right ventricle at the lateral, apical and/or basal levels in the vertical and/or horizontal axes. The 12 lead configuration (10 electrodes) is preferred for ischemia localization assessment, but 3, 5 or 7 lead configurations may still be used to provide a reasonable estimate of the location of cardiac ischemia.
6. Conditions under which an Abrupt Stop Exercise Protocol can be Considered as a Quasistationary Protocol
In general, each stage of gradually increasing and decreasing quasistationary exercise protocol is at least 3, 5, 8, or 10 minutes in duration. Each stage's duration is approximately an order of magnitude longer than the average duration (˜1 minute) of heart rate (HR) adjustment during an abrupt stop of the exercise between average peak load rate (˜120150 beat/min) and average rest (˜5070 beat/min) heart rate values.
One should note that the ˜1 minute HR adjustment period after an abrupt stop of exercise is typical only for healthy individuals and those with relatively moderate level of coronary artery disease. However, due to rapid development of massive exerciseinduced ischemia for the individuals with a pronounced coronary artery disease level the same adjustment period can be significantly longer—up to 10 minutes or more in duration. In these cases the HR deviations (described in the example 9 below) are small and indicate that such an ill patient remains under a significant physical stress even after an abrupt stop of exercise, without further exercise load exposure. In such cases a physician must usually interrupt the protocol prior to a completion of a full quasistationary exercise regiment. However, under these conditions a recovery portion of the QT/RR hysteresis loop, as well as a whole loop formed after an abrupt stop of exercise, can be considered as the quasistationary loop. Indeed, a slow heart rate recovery to a prior to the exercise HR level takes 3, 5, 8, 10 or more minutes in duration with small HR deviations and, therefore, still satisfies the underlying definition of a gradual exercise protocol (see example 9, below).
Thus, for very ill patients afflicted with coronary artery disease, an abrupt stop of exercise does not prevent the completion of the method of the present invention, since due to a prolonged HR recovery stage it still can be calculated as a quasistationary loop index without a completion of a full exercise protocol.
Typically a patient with a distinguished coronary artery disease level as determined by physical exhaustion, shortness of breath, chest pain, and/or some other clinical symptom, is unable to exercise longer than 3, 5, 8 or 10 or more minutes at even a lowlevel power output of about 20 watts. A gradual recovery process followed the abrupt stop of such exercise in such patients satisfies the definition of a quasistationary process as given herein (see, e.g., Example 9). Note that patients with moderate levels of coronary artery disease can exercise longer, up to 20 minutes or more, and may be exposed to significantly higher exercise workouts ranging from 50 to 300 watts.
7. RRinterval Monitoring with Blood Pressure and/or Pulse Signals
A quasistationary RR data set can be collected noninvasively not only via measurements of a cardiac surface ECG but also by monitoring a blood pressure and/or pulse signals. In these cases, instead of the ECG recorder, a system for assessing cardiac ischemia may comprise pulse and/or blood pressure monitors, as discussed below.
A pulse monitor or pulse meter may be suitable device, including but not limited to optoelectronic and phono or audio transducers attached to different parts of a subject's body (for example, to a finger as in finger plethysmography), which device measures a heart rate or pulse (HR). Preferably the device then computes RRintervals equal to 1/HR and stores these data in a computer memory in order to provide further RRinterval computational analysis as described herein.
A blood pressure monitor (e.g., a sphygmomanometer) can be any suitable device, including but not limited to a cuff, a stethoscope, or an automatic pressure registering system with a digital data storage module. Many such monitoring devices are applicable even for home use and typically contain all of the modules in one unit. An automatic cuff inflation monitor may also be included in the unit. Most units are portable and have a Dring cuff for onehanded application. The cuff usually fits around the upper arm or the wrist. These units provide personalized cuff inflation and deflation. They automatically adjust to changes in subject's blood pressure. Blood pressure monitoring with simultaneous measurement of the HR is convenient, easy to do and takes less than a minute per measurement. As in the case above, the RRintervals are equal to 1/HR. Such apparatus may be easily incorporated into a method and apparatus of the present invention with suitable interfaces, in accordance with known techniques.
The options described above may be used separately or in parallel, including in parallel with ECG data, depending on experimental needs. Pulse meter ischemia assessment is expected to be more accurate than blood pressure monitoring since an RR sampling frequency (frequency of HR measurements) for a pulse meter is at least an order of magnitude higher (10 to 1 data point) than in the case of blood pressure monitoring by an automatic sphygmomanometer.
The present invention is explained in greater detail in the nonlimiting examples set forth below.
A testing apparatus consistent with FIG. 3 was assembled. The electrocardiograms are recorded by an RZ152PM12 Digital ECG Holter Recorder (ROZINN ELECTRONICS, INC.; 7122 Myrtle Av., Glendale, N.Y., USA 113857254), via 12 electrical leads with LeadLok Holter/Stress Test Electrodes LL510 (LEADLOK, INC.; 500 Airport Way, P.O.Box L, Sandpoint, Id., USA 83864) placed on a subject's body in accordance with the manufacturer's instructions. Digital ECG data are transferred to a personal computer (Dell Dimension XPS T500 MHz/Windows 98) using a 40 MB flash card (RZFC40) with a PC 700 flash card reader, both from Rozinn Electronics, Inc. Holter for Windows (4.0.25) waveform analysis software is installed in the computer, which is used to process data by a standard computer based waveform analyzer software. Composite dispersionrestitution curves and an indicium that provides a quantitative characteristic of the extent of cardiac ischemia are then calculated manually or automatically in the computer through a program implemented in Fortran 90.
Experimental data were collected during an exercise protocol programmed in a Landice L7 Executive Treadmill (Landice Treadmills; 111 Canfield Av., Randolph, N.J. 07869). The programmed protocol included 20 stepwise intervals of a constant exercise load from 48 seconds to 1.5 minutes each in duration. Altogether these intervals formed two equalinduration gradually increasing and gradually decreasing exercise load stages, with total duration varying from 16 to 30 minutes. For each stage a treadmill belt speed and elevation varied thereandback, depending on the subject'sage and health conditions, from 1.5 miles per hour to 5.5 miles per hour and from one to ten degrees of treadmill elevation, respectively.
These examples illustrate quasistationary ischemiainduced QTRR interval hystereses in a variety of different human subjects. These data demonstrate a high sensitivity and the high resolution of the method.
These examples were carried out on two male subjects with an apparatus and procedure as described in Example 1 above. Referring to FIG. 5, one can readily see a significant difference in areas of hystereses between two generally healthy male subjects of different ages. These subjects (23 and 47 years old) exercised on a treadmill according to a quasistationary 30minute protocol with gradually increasing and gradually decreasing exercise load. Here squares and circles (thick line) indicate a hysteresis loop for the 23 year old subject, and diamonds and triangles (thin line) correspond to a larger loop for the 47 year old subject. Fitting curves are obtained using the thirdorder polynomial functions. A beat sampling rate with which a waveform analyzer determines QT and RR intervals is equal to one sample per minute. Neither of the subjects had a conventional ischemiainduced depression of the ECGST segments. However, the method of the present invention allows one to observe ischemiainduced hystereses that provide a satisfactory resolution within a conventionally subthreshold range of ischemic events and allows one to quantitatively differentiate between the hystereses of the two subjects.
These examples were carried out on two 55yearold male subjects with an apparatus and procedure as described in Example 1 above. FIG. 6 illustrates quasistationary QTRR interval hystereses for the male subjects. The curves fitted to the squares and empty circles relate to the first individual and illustrate a case of cardiac ischemia also detectable by the conventional ECG—ST segment depression technique. The curves fitted to the diamonds and triangles relate to the other subject, an individual who previously had experienced a myocardial infarction. These subjects exercised on a treadmill according to a quasistationary 20minute protocol with a gradually increasing and gradually decreasing exercise load. Fitting curves are obtained using thirdorder polynomial functions. These cases demonstrate that the method of the present invention allows one to resolve and quantitatively characterize the difference between (1) levels of ischemia that can be detected by the conventional ST depression method, and (2) low levels of ischemia (illustrated in FIG. 5) that are subthreshold for the conventional method and therefore undetectable by it. The levels of exerciseinduced ischemia reported in FIG. 5 are significantly lower than those shown in FIG. 6. This fact illustrates insufficient resolution of a conventional ST depression method in comparison with the method of the present invention.
This example was carried out with an apparatus and procedure as described in Example 1 above. FIG. 7 provides examples of quasistationary hystereses for a 55 yearold male subject before and after he engaged in a practice of regular aerobic exercise. Both experiments were performed according to the same quasistationary 20minute protocol with a gradually increasing and gradually decreasing exercise load. Fitting curves are obtained using thirdorder polynomial functions. The first test shows a pronounced exerciseinduced cardiac ischemic event developed near the peak level of exercise load, detected by both the method of the present invention and a conventional ECGST depression method. The maximum heart rate reached during the first test (before a regular exercise regimen was undertaken) was equal to 146. After a course of regular exercise the subject improved his cardiovascular health, which can be conventionally roughly, qualitatively, estimated by a comparison of peak heart rates. Indeed, the maximum heart rate at the peak exercise load from the first experiment to the second decreased by 16.4%, declining from 146 to 122. A conventional ST segment method also indicates the absence of ST depression, but did not provide any quantification of such an improvement since this ischemic range is subthreshold for the method. Unlike such a conventional method, the method of the present invention did provide such quantification. Applying the current invention, the curves in FIG. 7 developed from the second experiment show that the area of a quasistationary, QTRR interval hysteresis decreased significantly from the first experiment, and such hysteresis loop indicated that some level of exerciseinduced ischemia still remained. A change in the shape of the observed composite dispersionrestitution curves also indicates an improvement since it changed from a flatter curve, similar to the flatter curves (with a lower excitability and a higher threshold, v_{r=}0.3 to 0.35) in FIG. 2, to a healthier (less ischemic) more convexshape curve, which is similar to the lower threshold curves (v_{r=}0.2 to 0.25) in FIG. 2. Thus, FIG. 7 demonstrates that, due to its high sensitivity and high resolution, the method of the present invention can be used in the assessment of delicate alterations in levels of cardiac ischemia, indicating changes of cardiovascular health when treated by a conventional cardiovascular intervention.
This example was carried out with the data obtained in Examples 26 above. FIG. 8 illustrates a comparative cardiovascular health analysis based on ischemia assessment by the method of the present invention. In this example an indicium of cardiovascular health (here designated the cardiac ischemia index and abbreviated “CII”) was designed, which was defined as a quasistationary QTRR interval hysteresis loop area, S, normalized by dividing it by the product (T_{RR,max}−T_{RR,min})(T_{QT,max}−T_{QT,min}). For each particular subject this factor corrects the area for individual differences in the actual ranges of QT and RR intervals occurring during the tests under the quasistationary treadmill exercise protocol. We determined minimum and maximum CII in a sample of fourteen exercise tests and derived a normalized index <CII>=(CII−CII_{min})/(CII_{max}−CII_{min}) varying from 0 to 1. Alterations of <CII> in different subjects show that the method of the present invention allows one to resolve and quantitatively characterize different levels of cardiac and cardiovascular health in a region in which the conventional ST depression method is subthreshold and is unable to detect any exerciseinduced ischemia. Thus, unlike a rough conventional STsegment depression ischemic evaluation, the method of the present invention offers much more accurate assessing and monitoring of small variations of cardiac ischemia and associated changes of cardiac or cardiovascular health.
FIG. 9 illustrates a typical rapid sympathetic/parasympathetic nervous and hormonal adjustment of the QT (panels A, C) and RR (panels B,D) intervals to an abrupt stop after 10 minutes of exercise with increasing exercise load. All panels depict temporal variations of QT/RR intervals obtained from the right precordial lead V3 of the 12lead multilead electrocardiogram. A sampling rate with which a waveform analyzer determined QT and RR intervals was equal to 15 samples per minute. A human subject (a 47 yearsold male) was at rest the first 10 minutes and then began to exercise with gradually (during 10 minutes) increasing exercise load (Panels A, B—to the left from the RR, QT minima). Then at the peak of the exercise load (heart rate about 120 beat/min) the subject stepped off the treadmill in order to initialize the fastest RR and QT interval's adaptation to a complete abrupt stop of the exercise load. He rested long enough (13 minutes) in order to insure that QT and RR intervals reached postexercise average stationary values. Panels C and D demonstrate that the fastest rate of change of QT and RR intervals occurred immediately after the abrupt stop of the exercise load. These rates are about 0.015 s/min for QT intervals while they vary from 0.28 s to 0.295 s and about 0.15 s/min for RR intervals while they grow from 0.45 s to 0.6 s. Based on the abovedescribed experiment, a definition for “rapid sympathoadrenal and hormonal transients” or “rapid autonomic nervous system and hormonal transients” may be given.
Rapid transients due to autonomic nervous system and hormonal control refer to the transients with the rate of 0.15 s/min for RR intervals, which corresponds to the heart rate's rate of change of about 25 beat/min, and 0.02 s/min for QT intervals or faster rates of change in RR/QT intervals in response to a significant abrupt change (stop or increase) in exercise load (or other cardiac stimulus). The significant abrupt changes in exercise load are defined here as the load variations which cause rapid variations in RR/QT intervals, comparable in size with the entire range from the exercise peak to the stationary average rest values.
FIG. 10 illustrates a typical slow (quasistationary) QT (panel A) and RR (panel B) interval adjustment measured during gradually increasing and gradually decreasing exercise load in a right precordial V3 lead of the 12 lead electrocardiogram recording. The sampling was 15 QT and RR intervals per minute. A male subject exercised during two consecutive 10 minute long stages of gradually increasing and gradually decreasing exercise load. Both QT and RR intervals gradually approached the minimal values at about a peak exercise load (peak heart rate 120 beat/min) and then gradually returned to levels that were slightly lower than their initial preexercise rest values. The evolution of QT and RR intervals was well approximated by exponential fitting curves shown in gray in panels A and B. The ranges for the QTRR interval, thereandback, time variations were 0.34 s0.27 s0.33 s (an average rate of change ˜0.005 s/min) and 0.79 s0.47 s0.67 s (an average rate of change ˜0.032 s/min or ˜6 beat/min) for QT and RR intervals, respectively. The standard rootmeansquare deviation, σ, of the observed QT and RR intervals, shown by black dots in both panels, from their exponential fits were on an order of magnitude smaller than the average difference between the corresponding peak and rest values during the entire test. These deviations were σ˜0.003 s for QT and σ˜0.03 sfor RR intervals, respectively. According to FIG. 9 (panels C, D) such small perturbations, when associated with abrupt heart rate changes due to physiological fluctuations or due to discontinuity in an exercise load, may develop and decay faster than in 10 s, the time that is 60 times shorter than the duration of one gradual (ascending or descending) stage of the exercise protocol. Such a significant difference between the amplitudes and time constants of the QT/RR interval gradual changes and abrupt heart rate fluctuations allows one to average these fluctuations over time and fit the QT/RR protocol duration dynamics by an appropriate smooth exponentiallike function with a high order of accuracy. A simultaneous fitting procedure (panels A, B) determines an algorithm of a parametrical time dependence elimination from both measured QT/RR data sets and allows one to consider QT interval for each exercise stage as a monotonic function.
Based on the abovedescribed experiment a definition for a gradual, or “quasistationary” exercise (or stimulation) protocol, can be quantitatively specified: A quasistationary exercise (or stimulation) protocol refers to two contiguous stages (each stage 3, 5, 8 or 10 minutes or longer in duration) of gradually increasing and gradually decreasing exercise loads or stimulation, such as:
1. Each stage's duration is approximately an order of magnitude (e.g., at least about ten times) longer than the average duration (˜1 minute) of a heart rate adjustment during an abrupt stop of the exercise between average peak load rate (˜120150 beat/min) and average rest (˜5070 beat/min) heart rate values.
2. The standard rootmeansquare deviations of the original QT/RR interval data set from their smooth and monotonic (for each stage) fits are of an order of magnitude (e.g., at least about ten times) smaller than the average differences between peak and rest QT/RR interval values measured during the entire exercise under the quasistationary protocol.
As shown above (FIG. 10) a gradual quasistationary protocol itself allows one to substantially eliminate abrupt time dependent fluctuations from measured QT/RR interval data sets because these fluctuations have short durations and small amplitudes. Their effect can be even further reduced by fitting each RR/QT interval data set corresponding to each stage with a monotonic function of time. As a result the fitted QT interval values during each exercise stage can be presented as a substantially monotonic and smooth function of the quasistationary varying RR interval value. Presented on the (RRinterval, QTinterval)plane this function gives rise to a loop, whose shape, area and other measures depend only weakly on the details of the quasistationary protocol, and is quite similar to the hysteresis loop presented in FIG. 2. Similar to a generic hysteresis loop, this loop can be considered as primarily representing electrical conduction properties of cardiac muscle.
It is well known that exerciseinduced ischemia alters conditions for cardiac electrical conduction. If a particular individual has an exerciseinduced ischemic event, then one can expect that the two experimental composite dispersionrestitution curves corresponding to the ascending and descending stages of the quasistationary protocol will be different and will form a specific quasistationary hysteresis loop. Since according to a quasistationary protocol the evolution of the average values of QT and RR intervals occurs quite slowly as compared with the rate of the transients due to sympathetic/parasympathetic and hormonal control, the hysteresis loop practically does not depend on the peculiarities of the transients. In that case such a hysteresis can provide an excellent measure of gradual ischemic exercise dependent changes in cardiac electrical conduction and can reflect cardiac health itself and cardiovascular system health in general.
It should be particularly emphasized that neither J. Sarma et al, supra, nor A. Krahn's et al. supra, report work based on collecting quasistationary dependences, which are similar to the QTinterval—RR interval dependence, since in fact both studies were designed for different purposes. To the contrary, they were intentionally based on non—quasistationary exercise protocols that contained an abrupt exercise stop at or near the peak of the exercise load. These protocols generated a different type of QT/RR interval hysteresis loop with a substantial presence of nonstationary sympathoadrenal transients (see FIG. 9 above). Thus these prior art examples did not and could not include data which would characterize gradual changes in the dispersion and restitution properties of cardiac electrical conduction, and therefore included no substantial indication that could be attributed to exerciseinduced ischemia.
The following Examples describe various specific embodiments for carrying out the processing, shown in FIG. 4B, where the software implemented at steps 44 b and 45 b performs the following major steps:
(i) Generates sufficiently smooth time dependent QT and RR data sets by averaging/filtering and fitting these averaged data by exponential or any other suitable functions;
(ii) Combines into pairs the points of RR and QT data sets that correspond to the same time instants, thereby generating smooth QT/RR curves for the ascending and descending heart rate stages (branches) on the (T_{RR},T_{QT})plane or a similar plane or its image in computer memory;
(iii) Closes the ends of the QT/RR curves to transform them into a closed loop and determines an area, S, or a similar measure of the domain confined by the said loop and computes an index to provide a quantitative characteristic of cardiovascular health in a human subject based on a measure of this domain.
Each of these three major steps may consist of several sub steps, as illustrated in each of the methods shown in FIG. 11, FIG. 13 and FIG. 15, FIG. 16 and as discussed in greater detail below.
1. Raw data averaging/filtering (box 1 in FIG. 11). The raw data set consists of two subsets {t_{RR} ^{i},T_{RR} ^{i}} i=1,2, . . . N_{RR}−1, and {t_{QT} ^{i},T_{QT} ^{i}}, i=1,2, . . . N_{QT−}1, where t_{x} ^{i }and T_{x} ^{i }denote the ith sampling time instant and the respective, RR or QT, interval duration (subsript x stands for RR or QT). In order to simplify the notation we shall omit the subscripts RR and QT when it is applicable to both subsamples. It is convenient to represent each data point as a twocomponent vector u_{i}=(t^{i},T^{i}). The filtering procedure in this example comprises a moving averaging of neighboring data points. We shall denote a moving overage over a set of adjacent points by angular brackets with a subscript indicating the number of points included in the averaging operator. Thus, the preliminary data filtering at this sub step is described by the equation
for each i=1,2, . . . N−1, where N is a number of data points in the corresponding (RR or QT) raw data set. Thus, the RR and QT interval durations and the corresponding sampling time points are identically averaged in order to preserve the onetoone correspondence between them. This procedure removes the high frequency noise present in the raw data. This preliminary smoothing can also be described in the frequency domain and alternatively achieved via appropriate lowpass filtering. All the following processing pertinent to this example will be done on the averaged data points <U_{i}>_{2 }and the angular brackets will be omitted to simplify the notation.
2. Preliminary estimation of the tcoordinates of the minima (box 2 in FIG. 11). The sub step consists of preliminary estimation of the time instants, t_{RR} ^{i} ^{ m }and t_{QT} ^{j} ^{ m }of the minima of the initially averaged RRinterval and QTintervaldata sets, respectively. The algorithm does a sequential sorting of the data sets choosing M data points corresponding to M least values of the RR or QTintervals and then averages the result. (In our examples, M=10.) First, the algorithm finds the uvector u_{i} _{ 1 }=(t^{i} ^{ 1 },T^{i} ^{ 1 }) corresponding to the shortest interval T^{i} ^{ 1 }. Next, this data point is removed from the data set {u_{i}} and the algorithm sorts the remaining data set and again finds a uvector u_{i} _{ 2 }=(t^{i} ^{ 2 },T^{i} ^{ 2 }) corresponding to the shortest interval T^{i} ^{ 2 }. The minimum point is again removed from the data sets and sorting is repeated over and over again, until the M^{th }uvector u_{i} _{ m }=(t^{i} ^{ m },T^{i} ^{ m }) corresponding to the shortest interval T^{i} ^{ m }is found. Next, the algorithm averages all M pairs and calculates the average minimum time coordinate {overscore (t)} defined as
Finally, the algorithm determines the sampling time instant t^{i} ^{ m }, which is nearest to {overscore (t)}. We thus arrive to t_{RR} ^{i} ^{ m }and t_{QT} ^{j} ^{ m }for RR and QT data sets, respectively.
3. The first correction of the coordinates of the preliminary minima (box 3 in FIG. 11). At this sub step the first correction to the minimum coordinates t_{RR} ^{i} ^{ m }and t_{QT} ^{j} ^{ m }is found. This part of the algorithm is based on the iterative exponential fitting of the Tcomponent of the initially filtered data set {u_{i}}={t^{i},T^{i}} by functions of the form
where t^{m }is the time instant of the corresponding minimum determined at the previous sub step (i.e., t^{m}=t_{RR} ^{i} ^{ m }for RRintervals and t^{m}=t_{QT} ^{j} ^{ m }for QTintervals). The fitting is done separately for the descending (t<t^{m}) and ascending (t>t^{m}) branches of u(t) with the same value of constants A and t^{m }and different values of β for both branches. The initial value of A is taken from the preliminary estimation at sub step 2: A=T^{i} ^{ m }and t^{m}=t^{i} ^{ m }((i.e., A=T_{RR} ^{i} ^{ m },t^{m}=t_{RR} ^{i} ^{ m }for RRintervals and A=T_{QT} ^{j} ^{ m }, t^{m }for QTintervals). The initial value of constant β for each branch and each data sub set is obtained from the requirement that the initial mean root square deviation σ=σ_{0 }of the entire corresponding branch from its fit by equation (10.3) is minimum. In each subsequent iteration cycle new values of constants A and β are determined for each branch. At the beginning of each iteration cycle a constant A is taken from the previous step and the value of β is numerically adjusted to minimize the deviation value, σ. Such iterations are repeated while the mean square deviation σ is becoming smaller and are stopped when σ reaches a minimum value, σ=σ_{min}. At the end of this sub step the algorithm outputs the corrected values A and β and the respective values of σ_{min} ^{RR }and σ_{min} ^{QT}.
4. The preliminary smooth curves (box 4 in FIG. 11). The sub step consists of calculating a series of moving averages over p consecutive points for each data subset {u} as follows
We shall refer to the quantity p as the width of the averaging window. The calculation is performed for different values of p, and then the optimum value of the width of the averaging window p=m is determined by minimizing the mean square deviation of the set {u_{i} ^{p}}={t_{i} ^{p},T_{i} ^{p}} from its fit by Eq. (10.3) with the values of parameters A, t^{m}, and β determined at sub step 3. Having performed this procedure for each component of the data set (RR and QT) we arrive to the preliminary smoothed data set {u_{i} ^{m}}={t_{i} ^{m},T_{i} ^{m}}, where
The number of points in such a data set is N_{m}=N−m+1, where N is the number of data points after the initial filtering at sub step 1.
5. Correction of the preliminary smooth curves and the second correction of the QT and RR minimal values and its coordinates (box 5 in FIG. 11). At this sub step we redefine the moving averages (using a smaller averaging window) in the vicinity of the minimum of quantity T (RR or QT interval). This is useful to avoid distortions of the sought (fitting) curve T=T(t) near its minimum. The algorithm first specifies all data points {t^{i},T^{i}} such that T^{i }lie between
These data points have the superscript values i from the following set
Let us denote by i_{0}εI and i_{q}εI the subscript values corresponding to the earliest and latest time instants among {t_{i}} (iεI). We thus set i_{0}=min{I} and i_{q}=max{I}. Denoting the number of points in such a set by q_{RR }for the RR data and by q_{QT}, for the QT data we determine the width, q, of the averaging window as the minimum of the two, q=min{q_{QT},q_{RR}}. Now we write the moving average for the data points in the vicinity of the minimum as
Such modified averaging is applied to all consecutive data points u_{i }with the subscript i ranging from i_{0 }to i_{q}. The final (smoothed) data set consists of the first i_{0}−1 data points defined by Eq.(10.4) with i=1,2, . . . i_{0}−1, q points defined by Eq.(10.7) with i=i_{0}, i_{0}+1, . . . , i_{q}, and N−i_{q}−m points defined again by Eq.(10.4) with i=i_{q}+1, i_{q}+2, . . . , N−m +1. This the final set {{overscore (u)}_{i}} can be presented as
At the end of sub step 10.5 the algorithm determines the final minimum values of QT and RR interval and the corresponding time instants. The algorithm sorts the smoothed data sets (10.8) and determines {overscore (U)}_{min}≡({overscore (t)}_{min},{overscore (T)}_{min}) corresponding to the minimum value of {overscore (T)}_{i}. This can be written as
6. Final fitting and the final smooth QT and RR curves (box 6 in FIG. 11).
At this sub step the parameters of functions representing final smooth QT and RR curves are found. First, each data set {{overscore (u)}_{i}} is split into two subsets {u_{−} ^{i}} and {u_{+} ^{i}} corresponding to the descending ({overscore (t)}_{i}≦{overscore (t)}_{i} _{ m }) and ascending ({overscore (t)}_{i}≧{overscore (t)}_{i} _{ m }) branch, respectively. We also shift the time origin to the minimum point ({overscore (t)}_{i} _{ m }) and change the axis direction of the descending branch. These redefined variables are thus given by:
Notice that the minimum point is included in both branches. Next we perform a linear regression by fitting each branch with a 4th order polynomial function in the following form
The linear term is not included into this expression since this function must have a minimum at t_{±}=0. The value of d is defined by
For computational convenience the variables u_{±} are transformed into z_{±}=T_{±}−d and then the coefficients a_{±}, b_{±} and c_{±} are determined from the condition that expression for the error
where summation is performed over all points of the corresponding branch, is minimized. We thus obtain four similar sets of linear algebraic equations—two for each of the u_{±} branches of the two RR and QT data sets. These equations have the form
All summations are performed over all points of the branch. Now the QT and RR interval time dependences curves can be presented as the follows
where
are the fourth order polynomials given by Eq.(10.11) and
and
are the time coordinates of the corresponding minima defined by Eq. (10.9).
Thus, formulas (10.10)(10.16) determine the final smooth QT and RR curves with minima defined by formula (10.9).
7. Final smooth hysteresis loop (box 7 in FIG. 11). At this substep the software first generates a dense (N+1)point timegrid τ_{k}=t_{Start}+k(t_{end}−t_{start})/N, where k=0,1,2, . . . , N (N=1000 in this example) and t_{start }and t_{end }are the actual values of time at the beginning and end of the measurements, respectively. Then it computes the values of the four functions T_{RR} ^{±}(τ−t_{min} ^{RR}) and T_{QT} ^{±}(τ−t_{min} ^{QT}) on the grid, which parametrically represents the final smooth curves (T_{RR} ^{±}(τ−t_{min} ^{RR}), T_{QT} ^{±}(τ−t_{min} ^{QT})) in computer memory for the ascending (+) and descending (−) branches, respectively. This procedure is a computational equivalent of the analytical elimination of time. Next, the software also plots these smooth curves on the (T_{RR},T_{QT})plane or on another, similar plane, such as (ƒ_{RR},T_{QT}), where ƒ=1/T_{RR }is the instantaneous heart rate. Finally, the algorithm adds to the curves a set of points representing a closing line, which connects the end point of the lower (descending) branch with the initial point on the of the upper (ascending) branch and, thus, generates a closed QT/RR hysteresis loop.
8. A measure of the domain bounded by the QT/RR hysteresis loop (box 8 in FIG. 11). At this sub step a measure of the domain inside the QT/RR hysteresis loop is computed by numerically evaluating the following integral (see definition above):
where Ω. is the domain on the (T_{RR},T_{QT})plane with the boundary formed by the closed hysteresis loop, and ρ(T_{RR},T_{QT}) is a nonnegative (weight) function. In this example we can take ρ(T_{RR},T_{QT})=1 so that S coincides with the area of domain Ω, or we set ρ=1/(T_{RR})2 so S coincides with the dimensionless area of the domain inside the hysteresis loop on the (ƒ,T_{QT})plane, where ƒ=1/T_{RR }is the heart rate.
1. Raw data averaging/filtering. The sub step (similar to 10.1) consists of raw data averaging (filtering) according to formula (10.4). This is performed for a preliminary set of values of the averaging window width, ρ.
2. Final moving average smoothing. The sub step includes subsequent final smoothing of the preliminary smoothed data represented by the set {u_{i} ^{ρ}} found at the previous step: as follows.
The window width m_{f }is varied numerically to achieve optimum smoothness and accuracy of the fit. The optimally averaged data points form smooth curves on the corresponding planes (FIGS. 12,14).
3. A measure of the domain inside the QT/RR hysteresis loop. The procedure on this sub step is as defined in Example 10, step 8, above.
FIG. 15 illustrates major steps in the data processing procedure involving our nonlinear transformations method. The first three stages for the RR and QT data sets are quite similar and each results in the computation of the fitted birdlike curves, T^{RR}=T^{RR}(t) and T^{QT}=T^{QT}(t), on a dense timegrid. Having computed both birdlike curves one actually completes the data processing procedure, because after appropriate synchronization, these two dependences parametrically along with a closing line represent the sought hysteresis loop on the (T^{RR}, T^{QT}), or similar, plane. We will describe our method in general terms equally applicable for the QT and RR interval data sets, and indicate the specific instants where there is a difference in the algorithm.
1. Preliminary data processing stage. Let {T_{k}}, k=1,2, , N, be a set of the measured RR or QT interval durations, and {t_{k}} be a set of the corresponding time instants, so that t_{1 }and t_{N }are the starting and ending time instants (segments) of the entire record. The first three similar stages (stages 1 through 3 in FIG. 15) are the preliminary stage, the secondary, nonlineartransformation stage, and the computational stage. A more detailed data flow chart for these stages is shown in FIG. 16. The preliminary stage, indicated by boxes 1 through 7 in FIG. 16, is a combination of traditional data processing methods and includes: smoothing (averaging) the data sets (1), determination of a region near the minimum (2), and fitting a quadratic parabola to the data in this region (3), checking consistency of the result (4), renormalizing and centering the data at the minimum (5), cutting off the data segments outside the exercise region and separating the ascending and descending branches (6), and finally filtering off a data segment in near the minimum (7). Let us discuss these steps in more detail.
Box 1 in FIG. 16, indicates the moving averaging and itself consists of two steps. First, we specify the initial value of the moving average procedure parameter m, as
where r(x) is the integer closest to x, and the values of N_{c }and m_{min }are chosen depending on the number N of data points and the amount of random fluctuations in the data (the size of the random component in the data). In our examples we have chosen N_{c}=100 and m_{min}=3. The value of m can be redefined iteratively at the later stages and its choice will be discussed therein. Next we compute the moving average for {t_{i}} and {T_{i}} data sets with the given averaging parameter m as follows:
The subscript m will be omitted below if m is fixed and no ambiguity can arise. The next step in our algorithm is the initial determination of a time interval on which the parabolic fit will be performed, (the parabolic fit interval). This step is represented in FIG. 16 by Box 2. We note that this data subset can be redefined at a later stage if a certain condition is not satisfied. In the current realization of our algorithm this region defined differently for
data sets. For the data set
the initial parabolic fit segment is defined as the data segment
with all sequential values of i between i_{1 }and i_{2 }where
and the integer parameters n_{min }and Δm are defined as follows. The number n_{min }determines the time instant t_{n} _{ min }among the original set of time instants which is the closest to the minimum on the averaged data set {<t_{i}>,<T_{i}>}. In other words, t_{n} _{ min }is the nearest to the average time instant <t_{M}>which corresponds to the minimum value of the average RRinterval
that is, the time instant with the subscript value M defined by the condition
The value of Δm is linked to the value of m by the condition Δm=2m. The link between Δn and m arises from the requirement that the algorithm is stable and consistent. The consistency is the requirement that the positions of the minimum of the curve obtained by moving averaging and by quadratic fitting were approximately the same. On the other hand, it is also important that the value of m is sufficiently small because quadratic fitting is done with the purpose to describe the data only locally in the immediate vicinity of the minimum.
For the data set {t_{i}.T_{i} ^{QT}} the initial parabolic fit segment is defined as the data subset that consists of all consecutive points belonging to the lower portion of the nonaveraged {T_{i} ^{QT}} data set. We thus define i_{1 }as the first (minimum) number i such that
where R is a parameter, 0<R<1, that determines the portion of the data to be fitted with the quadratic parabola. In our calculations we set R=⅛=0.125, so we fit with the parabola the data points that correspond to the bottom 12.5% of the interval durations.
Thus, the subscript i_{1 }is the first (smallest) value of i such that condition (12.4) is satisfied. Similarly, i_{2 }is the last (greatest) value of i such that condition (12.4) is satisfied. The initial fit region is then defined as the following nonaveraged data subset:
This method can also be used for determining an initial data segment for the quadratic polynomial fitting of the RRdata set.
At the next step, shown by box 3 in FIG. 12.2, we fit the data in this region (for each data set) by a parabola so that the data are approximately represented by the equation
This is done using usual linear regression on this data subset. At the next step (Box 4 in FIG. 16) we check if the parabola has a minimum (i.e., if P_{1}>0). If this condition is satisfied, the determination of the parabolic fit interval and the fitting parabola itself is completed. Otherwise, we enter the loop indicated by boxes 1 a through 3 a in FIG. 16. The first step there is similar to the above second step utilized for the RR interval data set and defines an extended segment for the parabolic fitting. This is done by replacing m with m+1 and using this new value of m to redefine Δm=2m and also calculate new averaged data sets in accordance with Eq.(12.2) as indicated by Box 1a in FIG. 16. This allows us to evaluate a new value of n_{min }and new values of i_{1}=n_{min}−Δm and i_{2}=n_{min}+Δm with new value of m (Box 2a). Then the parabolic fit is done again (Box 3a) and the condition P_{1}>0, that the parabola has a minimum, is checked again. If it is satisfied, the determination of the fit interval and the quadratic fit coefficients procedure is completed. Otherwise we replace m with m+1 and repeat the process again and again until the condition P_{1}>0 is satisfied. This condition ensures that the extremum of the quadratic parabola is a minimum indeed. By the very design of the exercise protocol the average heart rate reaches a certain maximum somewhere inside the load stage and therefore corresponding average RRinterval reaches a minimum. This ensures that a sodefined data segment exists and is unique, and therefore the coefficients of the quadratic parabola are well defined. In short, we use the shortest data segment that is centered around the minimum of the averaged data set and that generates a quadratic parabola with a minimum (P_{1}>0).
The parabolic fit defines two important parameters of the data processing procedure, the position (t_{min},T_{min}) of the minimum on the (t,T)plane as follows
These parameters are final in the sense that our final fit curve will always pass through and have a minimum at the point (t_{min},T_{min}). The coordinates (t_{min},T_{min}) thus constitute parameters of our final fit birdlike curve. Having found the ordinate T_{min }of the minimum we can renormalization of the Tdata set as follows:
We also take the abscissa of the parabola's minimum t_{min }as the time origin and define the time components of the data points as follows
These two data transformations are indicated by Box 5 in FIG. 16. We also restrict the data set only to the exercise period, plus some short preceding and following intervals (Box 5). The conditions t_{i} ^{−}≦0, and t_{j} ^{+}>0 define the descending, {t_{i} ^{−},T_{i}}, and ascending, {t_{j} ^{+},T_{j}}, branches, respectively, so the corresponding data points can be readily identified and separated (Box 6). Given the durations t_{d }and t_{a }of the descending and ascending load stage, respectively, we can reduce the original set by cutting off the points on the descending branch with t_{i} ^{−}<−t_{d }and the points on the ascending branch with t_{j} ^{+>t} _{a }(Box 6 in FIG. 16). This determines the minimum value i_{0 }of subscript i for the descending branch, and the maximum value j_{max }of subscript j on the ascending branch.
The final step of the preliminary data processing is the conditional sorting (Box 7). The conditional sorting removes all consecutive points such that at least one of them falls below the minimum of the parabola. The removal of the points below the minimum is necessary because the nonlinear transformation is possible only when y_{i}≧1. Eliminating only separate points below the minimum y=1 would create a bias in the data. Therefore, we have to eliminate an entire segment of the data in the vicinity of the minimum. It should be remembered though, that these points have already been taken into account in the above quadratic filtering procedure that determines (t_{min},T_{min}) via Eq. (12.6).
Thus, the preliminary data processing results in two data sets corresponding to the descending and ascending load stages. The descending data set is defined as a set of consecutive pairs (t_{i} ^{−},y_{i})≡(t_{i} ^{−},y_{i} ^{−}) with i=i_{0}, i_{0}+1, i_{0}+2, . . . , i_{max}, where i_{max }is determined by the conditional sorting as the largest subscript i value still satisfying the condition y_{i}≧1. Similarly, the ascending data set is defined as a set of consecutive pairs (t_{i} ^{+},y_{i})≡(t_{i} ^{+},y_{i} ^{+}) with j=j_{min}, j_{0}+1, j_{0}+2, j_{max}, j_{min }is determined by the conditional sorting as the first subscript j value on the ascending branch starting from which the condition y_{j}≧1 is satisfied for all subsequent j.
2. Secondary data processing. At the second stage we introduce a fundamentally new method of nonlinear regression by means of two consecutive optimal nonlinear transformations. The idea of the method is to introduce for each branch two appropriate nonlinear transformations of the dependent and independent variables, y=ƒ(u) and u=φ(t), both transformations depending on some parameters, and choose the values of the parameters in such a way that the composition of these transformations ƒ(φ(t_{i})) would provide an approximation for y_{i}. Let (t_{i} ^{−}, y_{i} ^{−}) and (t_{i} ^{+}, y_{i} ^{+}) be the cut, conditionally sorted and normalized data set corresponding to the descending and ascending branches representing the dynamics of the RR or QTintervals during exercise. The nonlinear transformation y→u is defined via a smooth function
y=ƒ _{γ}(u), (12.9)
that has a unit minimum at u=1, so that ƒ_{γ}(1)=1, ƒ_{γ}′(1)=0, and ƒ_{γ}(u) grows monotonously when u≧1 and decreases monotonously when u≦0. The subscript yrepresents a set of discrete or continuous parameters and indicates a particular choice of such a function. Let us denote by ƒ_{γ} ^{−}(u) the monotonously decreasing branch of ƒ_{γ}(u) and its monotonously increasing branch, by ƒ_{γ} ^{+}(u). We can thus write
Let u=g_{γ} ^{+}(y) and u=g_{γ} ^{−}(y) be the inverse functions for the respective branches of ƒ_{γ}(u). The functions g_{γ} ^{+}(y) and ƒ_{γ} ^{+}(u) are monotonously increasing, while the functions g_{γ} ^{−}(y) and ƒ_{γ} ^{−}(u) are monotonously decreasing. Let {t_{i},y_{i} ^{−}} represent the data for the descending segment of the data set, i.e. t_{i}<t_{min}, and {t_{i},y_{i} ^{+}} represent the data for the ascending segment of the data set i.e. for t_{i}>t_{min}. The transformation
maps the monotonously decreasing (on the average) data set {t_{i},y_{i} ^{−}} into a monotonously increasing one:
Moreover, the average slope of the original data set is decreasing as t_{i }approaches t_{min }and eventually vanishing at the minimum. In contrast, the average slope of thetransformed data set is always nonzero at t═t_{min}. Similarly, the transformation
u _{γ,j} ^{+} =g _{γ} ^{+}(y_{j} ^{+}) (12.13)
maps the monotonously increasing (on the average) data set {t_{i},y_{i} ^{+}} into a monotonously increasing one:
In our examples we used a discrete parameter γ taking two values γ=1 and γ=2 that correspond to two particular choices of the nonlinear ytransformation. The first case (γ=1) is described by the equations
The second case, γ=2, is described by the equations
Both functions reach a minimum y=1 at u=1. An example of such a function depending on a continuous parameter γ>0 is given by
The parameter b_{γ} has been determined from the condition that ƒγ(u) has a minimum at u=1 and the coefficient A_{γ} is determined by the condition ƒ_{γ}(1)=1, which yields
In our numerical examples below we utilize the discrete parameter case with γ taking two values, 1 and 2. The original and transformed sets are shown in FIGS. 17 (RR intervals) and 18 (QTintervals). Panels A in both figures show the original data sets on the (t,T)plane and panels B and C show the transformed sets on the (t,u)plane, for γ=1 and for γ=2, respectively. The parabola minima on panels A and C are marked with a circled asterisk. The original data points concentrate near a nonmonotonous curve (the average curve). The data points on the transformed plane concentrate around a monotonously growing (average) curve. Moreover, the figure illustrates that the transformation changes the slope of the average curve in the vicinity of t=t_{min }from zero to a finite, nonzero value. On the (t−t_{min},u)plane the point corresponding to the parabola's minimum is (0,1). This point is also marked by a circled asterisk on panels B and C of FIGS. 17 and 18.
Let us introduce a pair of new time variables τ^{−} and τ^{+} for the descending and ascending branches, respectively. We shall count them off from the abscissa of the minimum of the fitted parabola and set
where I^{−} and J^{+} are the number of data points on the descending and ascending branches, respectively. The (formerly) descending branch can be treated in exactly the same way as the ascending one if simultaneously with the time inversion given by the first line in Eq. (12.19) we perform an additional transformation of the descending branch ordinates as follows
In these variables both branches (τ_{i} ^{−},v_{i}) and (τ_{j} ^{+},u_{j} ^{+}) are monotonously growing and start from the same point (τ=0,u=1) at which both have nonzero slope and possess similar behavior (convexity).
Since both branches are treated in exactly the same way, we shall simplify the notation and temporarily omit the superscripts ± and write (τ_{k},u_{k}) for any of the pairs (τ_{i} ^{−},v_{i}) or (τ_{j} ^{+},u_{j} ^{+}). We shall fit the data set {u_{k}} with {φ(α, β, τ_{k})}, that is represent {u_{k}} as
where the function φ depends linearly on K and is defined as
where
A family of functions ξ(α,β,τ) is shown in FIG. 19 for fifteen values of parameter β. The parameter α is completely scaled out by plotting the function on the plane (τ/αε/α). The function ξ(α,β,τ) is continuous in all three variables and as a function of τ at fixed α and β has a unit slope at τ=0, ξ′(α,β,0)=1. Therefore, at τ→0 the function ξ possesses a very simple behavior, ξ˜τ, which is independent of α and β (the size of the region of such behavior of course depends on α and β). The function ξpossesses the following important feature: when β passes through the point β=0, its asymptotic behavior at τ→∞ continuously changes from the power function, ξ˜τ^{β} at β>0, through ξ˜ln(τ) at β=0, to a power of the logarithm, ξ˜ln^{1+β}(τ) when −1<β<0. When β→−1, the behavior of ξ changes once more, and becomes ξ˜ln(ln(τ)). The convexity of any function of the family is the same when β<1.
We shall explicitly express parameter K via α and β using the requirement for a given pair of α and β Eqs. (12.22) and (12.23) ensure the best data fit in the uspace in a certain vicinity of the point (τ=0,u=1). Let K_{1 }be the number of points where the fit is required. The corresponding quadratic error is then a function of K, α and β that can be written as
and the requirement of its minimum immediately yields the following expression for K
In our calculations we used such K_{1 }value that would include all adjacent points with the values of u between u=1 and u=1+0.4 (u_{max}−1). We have thus reduced the number of fitting parameters in our fitting procedure to two continuous parameters, α and β, and one discrete parameter, γ. The fitting function thus becomes
The values of parameters α, β and K (and γ) are now directly determined by the condition that the fit error in the yspace
is minimum. The sum in Eq. (12.27) is evaluated numerically on a grid (α,β) values for γ equal 1 and 2. Then the values of α, β and γ that deliver a minimum to ε_{γ} ^{(y) }are found via numerical trials. Calculations can then be repeated on a finer grid in the vicinity of the found minimum.
Having found parameters α_{min}, β_{min }and γ_{min }for the ascending branch we generate a dense tgrid {t_{s}, s=1,2, . . . , N} and calculate the corresponding values of the interval duration as
A similar dense representation of the descending branch can be calculated in exactly the same way. The resulting birdlike curves are illustrated in panels A and C of FIG. 20. The actual absolute and relative errors are indicated in the captions. The right hand side panels B and D represent hysteresis curves in two different representations and each resulting from the curves shown in Panels A and C. The following computations of the hysteresis loop and its measure given by Eq. (10.17) is then performed as described in Example 10.
In addition to the procedures generating a hysteresis loop on the plane (QTinterval versus RRinterval) or an equivalent plane, one can introduce and assess a separate hysteresis of the duration of RRinterval versus exercise workload, which gradually varies, thereandback, during the ascending and descending exercise stages. The RRhysteresis can be displayed as loops on different planes based on just a single RR data set {t_{RR} ^{i},T_{RR} ^{i}} analysis. For example, such a loop can be displayed on the (τ^{i},T_{RR} ^{i})plane, where τ^{i}=t_{RR} ^{i}−t_{min} and t_{min }is the time instant corresponding to the peak of the exercise load, or the center of the maximum load period, which may be determined according to numerical techniques described in the examples 1012. The RR loop can also be introduced on the (W(t_{RR} ^{i}),T_{RR} ^{i}) or (τ^{i},(T_{RR} ^{i})^{−1}) planes.
Here W(t_{RR} ^{i}) is a workload that varies versus exercise stages, i.e. time and (T_{RR} ^{i})^{−l }represents the heart rate.
In order to apply a numerical technique from the Example 10 (or any other as in Examples 11 or 12), one repeats essentially the same computational steps described in these examples. However, instead of considering both QT {t_{QT} ^{i},T_{QT} ^{i}} and RR {t_{RR} ^{i},T_{RR} ^{j}} interval data sets, only a single RR data set is numerically processed throughout the whole sequence of the described stages for a creation of a hysteresis loop. In this case variables τ^{i},(T_{RR} ^{i})^{−1 }or W(t_{RR} ^{i}) play the role of the second alternating component that along with the first T_{RR} ^{i }variable forms the RRhysteresis plane.
An apparatus for assessing cardiac ischemia incorporating heart rate feedback is schematically illustrated in FIG. 14. The apparatus comprises an exercise apparatus 102 such as a treadmill, stationary cycle, rowing machine or the like, which exercise apparatus includes an adjustable exercise load feature. A heart rate monitor 103 is connected to a subject 101 by line 104, or by other suitable means such as a telemetry or radio transmitter device. The heart rate monitor is connected to a controller 105 by line 106, which controller is connected back to the exercise apparatus by line 107. The controller may be any suitable hardware, software, or hardware/software device, such as a personal computer equipped with appropriate input/output interface boards. The controller is configured to, first, gradually increase the load of the exercise apparatus (i.e., the exercise load imposed on the subject) in response to heart rate data from the subject to produce a stage of gradually increasing heart rate in the subject, and then gradually decrease the load of the exercise apparatus in response to heart rate data from the subject to produce a stage of gradually decreasing heart rate in said subject. In the alternative, in suitable subjects afflicted with coronary artery disease, the stage of gradually decreasing heart rate may occur after an abrupt stop as described above, with data collected from the subject after the abrupt stop.
The controller 105 may optionally include hardware, software, or hardware and software as described above for collecting a first RRinterval data set (e.g., a first QT and RRinterval data set) from the subject during the stage of gradually increasing heart rate, and for collecting a second RRinterval data (e.g., a second QT and RRinterval data set) set from the subject during the stage of gradually decreasing heart rate. In the alternative, such data may be collected by a separate apparatus, as also described above.
The foregoing examples are illustrative of the present invention and are not to be construed as limiting thereof. The invention is defined by the following claims, with equivalents of the claims to be included therein.
Claims (8)
Priority Applications (3)
Application Number  Priority Date  Filing Date  Title 

US09/603,286 US6361503B1 (en)  20000626  20000626  Method and system for evaluating cardiac ischemia 
US09/891,910 US6652467B2 (en)  20000626  20010626  Method and system for evaluating cardiac ischemia with RRinterval data sets 
US10/036,004 US6768919B2 (en)  20000626  20011226  Method and system for evaluating cardiac ischemia with heart rate feedback 
Applications Claiming Priority (1)
Application Number  Priority Date  Filing Date  Title 

US10/036,004 US6768919B2 (en)  20000626  20011226  Method and system for evaluating cardiac ischemia with heart rate feedback 
Related Parent Applications (1)
Application Number  Title  Priority Date  Filing Date  

US09/891,910 ContinuationInPart US6652467B2 (en)  20000626  20010626  Method and system for evaluating cardiac ischemia with RRinterval data sets 
Publications (2)
Publication Number  Publication Date 

US20020151811A1 US20020151811A1 (en)  20021017 
US6768919B2 true US6768919B2 (en)  20040727 
Family
ID=46204372
Family Applications (1)
Application Number  Title  Priority Date  Filing Date 

US10/036,004 Expired  Fee Related US6768919B2 (en)  20000626  20011226  Method and system for evaluating cardiac ischemia with heart rate feedback 
Country Status (1)
Country  Link 

US (1)  US6768919B2 (en) 
Cited By (17)
Publication number  Priority date  Publication date  Assignee  Title 

US20030149370A1 (en) *  20000626  20030807  Starobin Joseph M.  Method and system for evaluating cardiac ischemia with an exercise protocol 
US20040021317A1 (en) *  20020731  20040205  Brockway Robert D.  Expandable band and locking mechanism 
US20040088017A1 (en) *  20021031  20040506  Vinod Sharma  Ischemia detection based on cardiac conduction time 
WO2005009234A1 (en) *  20030723  20050203  Mediwave Star Technology, Inc.  Method and system for evaluating cardiac ischemia based on heart rate fluctuations 
US6904313B1 (en) *  20020523  20050607  Pacesetter, Inc.  Methods and systems for monitoring heart rate recovery 
US6969382B2 (en)  19990726  20051129  Zuli Holdings, Ltd.  Method and apparatus for treating bodily tissues with medicinal substance 
WO2006041337A1 (en) *  20041011  20060420  St Jude Medical  Detection of ischemia 
US20060116611A1 (en) *  19990726  20060601  Jacob Richter  Apparatus and method for treating body tissues with electricity or medicaments 
US20070021678A1 (en) *  20050719  20070125  Cardiac Pacemakers, Inc.  Methods and apparatus for monitoring physiological responses to steady state activity 
US20070054871A1 (en) *  20050906  20070308  Pastore Joseph M  Method and apparatus for device controlled gene expression for cardiac protection 
US20070250127A1 (en) *  20060419  20071025  Lee Stylos  Implantable Medical Device with Automatic Ischemia Threshold Determination 
US20080081354A1 (en) *  20061002  20080403  Cardiac Pacemakers, Inc.  Devices, vectors and methods for inducible ischemia cardioprotection 
US20100259407A1 (en) *  20090408  20101014  Polar Electro Oy  Portable Apparatus 
US20100312130A1 (en) *  20060627  20101209  Yi Zhang  Graded response to myocardial ischemia 
US8000780B2 (en)  20060627  20110816  Cardiac Pacemakers, Inc.  Detection of myocardial ischemia from the time sequence of implanted sensor measurements 
US20110201943A1 (en) *  20100216  20110818  Beck Kenneth C  Kinetics of physiological response to activity during activities of daily living 
US9241677B2 (en)  20130702  20160126  MobHealth Corporation  Device and methods for assessing, diagnosing, and/or monitoring heart health 
Families Citing this family (28)
Publication number  Priority date  Publication date  Assignee  Title 

US7107093B2 (en) *  20030429  20060912  Medtronic, Inc.  Use of activation and recovery times and dispersions to monitor heart failure status and arrhythmia risk 
WO2005006946A2 (en) *  20030703  20050127  New York Univeristy  System and method for assessment of cardiac electrophysiologic stability and modulation of cardiac oscillations 
US7031766B1 (en)  20040420  20060418  Pacesetter, Inc.  Methods and devices for determining exercise diagnostic parameters 
US7676262B1 (en)  20040420  20100309  Pacesetter, Inc.  Methods and devices for determining exercise compliance diagnostics 
US7043294B1 (en)  20040420  20060509  Pacesetter, Inc.  Methods and devices for determining heart rate recovery 
US20060122651A1 (en) *  20041203  20060608  Whitman Teresa A  Use of mechanical restitution to predict hemodynamic response to a rapid ventricular rhythm 
US7297114B2 (en) *  20050125  20071120  Pacesetter, Inc.  System and method for distinguishing among cardiac ischemia, hypoglycemia and hyperglycemia using an implantable medical device 
DE102006058332A1 (en) *  20061211  20080619  Siemens Ag  A method for controlling a measurement of a magnetic resonance apparatus with reference to an electrocardiogram signal 
US9320448B2 (en)  20080418  20160426  Pacesetter, Inc.  Systems and methods for improved atrial fibrillation (AF) monitoring 
DE102008039917A1 (en) *  20080827  20100304  Biosigna Gmbh  Patient's electrocardiogram analysis arrangement for use in e.g. electrocardiogram device, has test unit testing whether measured value of parameter corresponds to normal value and/or lies within or outside of normal value range 
US9314209B2 (en) *  20120911  20160419  Covidien Lp  Methods and systems for determining physiological information based on a correlation sequence 
US9848828B2 (en)  20131024  20171226  Logitech Europe, S.A.  System and method for identifying fatigue sources 
US9626478B2 (en)  20131024  20170418  Logitech Europe, S.A.  System and method for tracking biological age over time based upon heart rate variability 
US10078734B2 (en)  20131024  20180918  Logitech Europe, S.A.  System and method for identifying performance days using earphones with biometric sensors 
US20150116125A1 (en) *  20131024  20150430  JayBird LLC  Wristband with removable activity monitoring device 
US9864843B2 (en)  20131024  20180109  Logitech Europe S.A.  System and method for identifying performance days 
US9314172B2 (en)  20131024  20160419  JayBird LLC  System and method for providing a training load schedule for peak performance positioning 
US9622685B2 (en)  20131024  20170418  Logitech Europe, S.A.  System and method for providing a training load schedule for peak performance positioning using earphones with biometric sensors 
US9849538B2 (en)  20141224  20171226  Logitech Europe, S.A.  Watertight welding methods and components 
USD777186S1 (en)  20141224  20170124  Logitech Europe, S.A.  Display screen or portion thereof with a graphical user interface 
USD784961S1 (en)  20150605  20170425  Logitech Europe, S.A.  Ear cushion 
US9729953B2 (en)  20150724  20170808  Logitech Europe S.A.  Wearable earbuds having a reduced tip dimension 
US9743745B2 (en)  20151002  20170829  Logitech Europe S.A.  Optimized cord clip 
US10117015B2 (en)  20151020  20181030  Logitech Europe, S.A.  Earphones optimized for users with small ear anatomy 
US10292606B2 (en)  20151105  20190521  Logitech Europe, S.A.  System and method for determining performance capacity 
US9986323B2 (en)  20151119  20180529  Logitech Europe, S.A.  Earphones with attachable expansion pack 
US10112075B2 (en)  20160201  20181030  Logitech Europe, S.A.  Systems, methods and devices for providing a personalized exercise program recommendation 
US10129628B2 (en)  20160201  20181113  Logitech Europe, S.A.  Systems, methods and devices for providing an exertion recommendation based on performance capacity 
Citations (19)
Publication number  Priority date  Publication date  Assignee  Title 

US4860763A (en) *  19870729  19890829  Schminke Kevin L  Cardiovascular conditioning and therapeutic system 
US4870974A (en)  19870930  19891003  Chinese Pla General Hospital  Apparatus and method for detecting heart characteristics by way of electrical stimulation 
US4976424A (en) *  19870825  19901211  Schwinn Bicycle Company  Load control for exercise device 
US5020540A (en)  19871009  19910604  Biometrak Corporation  Cardiac biopotential analysis system and method 
US5117834A (en)  19900806  19920602  Kroll Mark W  Method and apparatus for noninvasively determing a patients susceptibility to ventricular arrhythmias 
US5148812A (en)  19910220  19920922  Georgetown University  Noninvasive dynamic tracking of cardiac vulnerability by analysis of twave alternans 
US5323783A (en)  19921112  19940628  Del Mar Avionics  Dynamic ST segment estimation and adjustment 
US5419338A (en)  19940222  19950530  City Of Hope  Autonomic nervous system testing by bivariate spectral analysis of heart period and QT interval variability 
US5560370A (en)  19910220  19961001  Georgetown University  Method and apparatus for prediction of cardiac electrical instability by simultaneous assessment of Twave alternans and QT interval dispersion 
US5570696A (en) *  19940126  19961105  Cambridge Heart, Inc.  Method and apparatus for assessing myocardial electrical stability 
US5713367A (en)  19940126  19980203  Cambridge Heart, Inc.  Measuring and assessing cardiac electrical stability 
US5738612A (en) *  19961204  19980414  Colin Corporation  Exercise apparatus having exerciseload changing function 
US5792065A (en)  19970318  19980811  Marquette Medical Systems, Inc.  Method and apparatus for determining Twave marker points during QT dispersion analysis 
US5794623A (en)  19960927  19980818  HewlettPackard Company  Intramyocardial Wenckebach activity detector 
US5827195A (en)  19970509  19981027  Cambridge Heart, Inc.  Electrocardiogram noise reduction using multidimensional filtering 
US5842997A (en)  19910220  19981201  Georgetown University  Noninvasive, dynamic tracking of cardiac vulnerability by simultaneous analysis of heart rate variability and Twave alternans 
US5891047A (en)  19970314  19990406  Cambridge Heart, Inc.  Detecting abnormal activation of heart 
US5951484A (en)  19970801  19990914  Harbinger Medical, Inc.  Method of noninvasively determining a patient's susceptibility to arrhythmia 
US6361503B1 (en) *  20000626  20020326  Mediwave Star Technology, Inc.  Method and system for evaluating cardiac ischemia 

2001
 20011226 US US10/036,004 patent/US6768919B2/en not_active Expired  Fee Related
Patent Citations (20)
Publication number  Priority date  Publication date  Assignee  Title 

US4860763A (en) *  19870729  19890829  Schminke Kevin L  Cardiovascular conditioning and therapeutic system 
US4976424A (en) *  19870825  19901211  Schwinn Bicycle Company  Load control for exercise device 
US4870974A (en)  19870930  19891003  Chinese Pla General Hospital  Apparatus and method for detecting heart characteristics by way of electrical stimulation 
US5020540A (en)  19871009  19910604  Biometrak Corporation  Cardiac biopotential analysis system and method 
US5117834A (en)  19900806  19920602  Kroll Mark W  Method and apparatus for noninvasively determing a patients susceptibility to ventricular arrhythmias 
US5842997A (en)  19910220  19981201  Georgetown University  Noninvasive, dynamic tracking of cardiac vulnerability by simultaneous analysis of heart rate variability and Twave alternans 
US5148812A (en)  19910220  19920922  Georgetown University  Noninvasive dynamic tracking of cardiac vulnerability by analysis of twave alternans 
US5921940A (en)  19910220  19990713  Georgetown University  Method and apparatus for using physiologic stress in assessing myocardial electrical stability 
US5560370A (en)  19910220  19961001  Georgetown University  Method and apparatus for prediction of cardiac electrical instability by simultaneous assessment of Twave alternans and QT interval dispersion 
US5323783A (en)  19921112  19940628  Del Mar Avionics  Dynamic ST segment estimation and adjustment 
US5570696A (en) *  19940126  19961105  Cambridge Heart, Inc.  Method and apparatus for assessing myocardial electrical stability 
US5713367A (en)  19940126  19980203  Cambridge Heart, Inc.  Measuring and assessing cardiac electrical stability 
US5419338A (en)  19940222  19950530  City Of Hope  Autonomic nervous system testing by bivariate spectral analysis of heart period and QT interval variability 
US5794623A (en)  19960927  19980818  HewlettPackard Company  Intramyocardial Wenckebach activity detector 
US5738612A (en) *  19961204  19980414  Colin Corporation  Exercise apparatus having exerciseload changing function 
US5891047A (en)  19970314  19990406  Cambridge Heart, Inc.  Detecting abnormal activation of heart 
US5792065A (en)  19970318  19980811  Marquette Medical Systems, Inc.  Method and apparatus for determining Twave marker points during QT dispersion analysis 
US5827195A (en)  19970509  19981027  Cambridge Heart, Inc.  Electrocardiogram noise reduction using multidimensional filtering 
US5951484A (en)  19970801  19990914  Harbinger Medical, Inc.  Method of noninvasively determining a patient's susceptibility to arrhythmia 
US6361503B1 (en) *  20000626  20020326  Mediwave Star Technology, Inc.  Method and system for evaluating cardiac ischemia 
NonPatent Citations (19)
Title 

Arnold et al.; The dependence on heart rate of the human ventricular action potential duration, Cardiovascular Research, 16, 547551 (1982). 
Chernyak et al.; Class of Exactly Solvable Models of Excitable Media, Phys. Rev. Lett,. 80:25, 56755678 (1998). 
Chernyak et al.; Where do dispersion curves end? A basic question in theory of excitable media, Phys. Rev. E,. 58:4, 41084111 (1998). 
Ciavolella et al.; Exponential Fit of QT IntervalHeart Rate Relation During Exercise Used to Diagnose Stressinduced Myocardial Ischemia, Journal of Electrocardiology, 24:2, 145153 (1991). 
Cole et al.; HeartRate Recovery Immediately After Exercise As A Predictor Of Mortality, The New England Journal of Medicine, 341:18, 13511357 (Oct. 1999). 
Franz et al.; Cycle Length Dependence of Human Action Potential Duration In Vivo; Effects of Single Extrastimuli, Sudden Sustained Rate Acceleration and Deceleration, and Different SteadyState Frequencies, J. Clin. Invest,. 82, 972979 (1988). 
Froelicher, Jr. et al.; A comparison of three maximal treadmill exercise protocols, Journal of Applied Physiology, 36:6, 720725 (1974). 
Hintze et al.; Prognostic Properties of QT/RR Dynamics in Survivors of Myocardial Infarction with Reduced Systolic Function, NASPE Annual Meeting, Washington, D.C. (May 1720, 2000). 
International Search Report, International Application No. PCT/US01/20391 dated Aug. 20, 2001. 
Jonnalegedda et al.; An Exponential Formula for Heart Rate Dependence of QT Interval During Exercise and Cardiac Pacing in Human: Reevaluation of Bazett's Formula, Am J Cardiol, 54, 103108 (1984). 
Jonnalegedda et al.; Hysteresis in the Human RRQT Relationship During Exercise and Recovery, PACE 10, 485491 (1997). 
Krahn, M.D. et al.; Hysteresis of the RT Interval With Exercise; A New Marker for the LongQT Syndrome?, Circulation, 96, 15511556 (1997). 
Lau et al.; Hysteresis of the ventricular paced QT interval in response to abrupt changes in pacing rate, Cardiovascualr Research, 22, 6772 (1998) 
Pierpoint et al.; Heart rate recovery postexercise as an index of parasympathetic activity, Journal of the Autonomic Nervous System, 80, 169174 (May 12, 2000). 
Sarma et al, "Hysteresis . . . Recovery", MayJun. 1987.* * 
Starobin et al.; The role of a critical excitation length scale in dynamics of reentrant cardiac arrhythmias, Herzschr Elektrophys, 10, 119136 (Month Unknown, 1999). 
Surawicz; Will QT Dispersion Play a Role in Clinical DecisionMaking?, J Cardiovascular Electrophysiol, 7, 777784 (1996) 
Swan et al.; Rate adaption of QT intervals during and after exercise in children witj congenital long QT syndrome, European Heart Journal, 19, 508513 (1998). 
Takahashi et al.; Paradoxically Shortened QT Interval after a Prolonged Pause, PACE, 21, 14761479 (1998). 
Cited By (30)
Publication number  Priority date  Publication date  Assignee  Title 

US6969382B2 (en)  19990726  20051129  Zuli Holdings, Ltd.  Method and apparatus for treating bodily tissues with medicinal substance 
US9289377B2 (en)  19990726  20160322  Microtech Medical Technologies Ltd.  Apparatus and method for treating body tissues with electricity or medicaments 
US7785319B2 (en)  19990726  20100831  Microtech Medical Technologies Ltd.  Method and apparatus for treating bodily tissues with medicinal substance 
US8108041B2 (en)  19990726  20120131  Zuli Holdings, Ltd.  Apparatus and method for treating body tissues with electricity or medicaments 
US20060116611A1 (en) *  19990726  20060601  Jacob Richter  Apparatus and method for treating body tissues with electricity or medicaments 
US20030149370A1 (en) *  20000626  20030807  Starobin Joseph M.  Method and system for evaluating cardiac ischemia with an exercise protocol 
US7104961B2 (en) *  20000626  20060912  Mediwave Star Technology, Inc.  Method and system for evaluating cardiac ischemia with an exercise protocol 
US6904313B1 (en) *  20020523  20050607  Pacesetter, Inc.  Methods and systems for monitoring heart rate recovery 
US20040021317A1 (en) *  20020731  20040205  Brockway Robert D.  Expandable band and locking mechanism 
US7415307B2 (en) *  20021031  20080819  Medtronic, Inc.  Ischemia detection based on cardiac conduction time 
US20040088017A1 (en) *  20021031  20040506  Vinod Sharma  Ischemia detection based on cardiac conduction time 
US20050038351A1 (en) *  20030723  20050217  Starobin Joseph M.  Method and system for evaluating cardiac ischemia based on heart rate fluctuations 
WO2005009234A1 (en) *  20030723  20050203  Mediwave Star Technology, Inc.  Method and system for evaluating cardiac ischemia based on heart rate fluctuations 
US7778697B2 (en)  20041011  20100817  St. Jude Medical Ab  Detection of ischemia 
WO2006041337A1 (en) *  20041011  20060420  St Jude Medical  Detection of ischemia 
US20090143690A1 (en) *  20041011  20090604  Anders Bjorling  Detection of ischemia 
US20070021678A1 (en) *  20050719  20070125  Cardiac Pacemakers, Inc.  Methods and apparatus for monitoring physiological responses to steady state activity 
US7774057B2 (en)  20050906  20100810  Cardiac Pacemakers, Inc.  Method and apparatus for device controlled gene expression for cardiac protection 
US20070054871A1 (en) *  20050906  20070308  Pastore Joseph M  Method and apparatus for device controlled gene expression for cardiac protection 
US8538520B2 (en)  20050906  20130917  Cardiac Pacemakers, Inc.  Method and apparatus for device controlled gene expression for cardiac protection 
US7996084B2 (en) *  20060419  20110809  Medtronic, Inc.  Implantable medical device with automatic ischemia threshold determination 
US20070250127A1 (en) *  20060419  20071025  Lee Stylos  Implantable Medical Device with Automatic Ischemia Threshold Determination 
US20100312130A1 (en) *  20060627  20101209  Yi Zhang  Graded response to myocardial ischemia 
US8000780B2 (en)  20060627  20110816  Cardiac Pacemakers, Inc.  Detection of myocardial ischemia from the time sequence of implanted sensor measurements 
US20080081354A1 (en) *  20061002  20080403  Cardiac Pacemakers, Inc.  Devices, vectors and methods for inducible ischemia cardioprotection 
US20100259407A1 (en) *  20090408  20101014  Polar Electro Oy  Portable Apparatus 
US8988241B2 (en) *  20090408  20150324  Polar Electro Oy  Portable apparatus 
US20110201943A1 (en) *  20100216  20110818  Beck Kenneth C  Kinetics of physiological response to activity during activities of daily living 
US9730618B2 (en)  20100216  20170815  Cardiac Pacemakers, Inc.  Kinetics of physiological response to activity during activities of daily living 
US9241677B2 (en)  20130702  20160126  MobHealth Corporation  Device and methods for assessing, diagnosing, and/or monitoring heart health 
Also Published As
Publication number  Publication date 

US20020151811A1 (en)  20021017 
Similar Documents
Publication  Publication Date  Title 

Winchell et al.  Spectral analysis of heart rate variability in the ICU: a measure of autonomic function  
Crawford et al.  ACC/AHA guidelines for ambulatory electrocardiography: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee to Revise the Guidelines for Ambulatory Electrocardiography) developed in collaboration with the North American Society for Pacing and Electrophysiology  
Lu et al.  A comparison of photoplethysmography and ECG recording to analyse heart rate variability in healthy subjects  
Kelly et al.  Effective arterial elastance as index of arterial vascular load in humans.  
EP0462690B1 (en)  An improved surface ECG frequency analysis system and method based upon spectral turbulence estimation  
US4870974A (en)  Apparatus and method for detecting heart characteristics by way of electrical stimulation  
US7043293B1 (en)  Method and apparatus for waveform assessment  
US6656125B2 (en)  System and process for analyzing a medical condition of a user  
US6648828B2 (en)  Continuous, noninvasive technique for measuring blood pressure using impedance plethysmography  
US6485429B2 (en)  Method and a device for noninvasive measurement of the blood pressure and for detection of arrhythmia  
US7846106B2 (en)  Atrial fibrillation detection using SPO2  
CA2422801C (en)  Noninvasive measurement of suprasystolic signals  
JP3242981B2 (en)  Method and apparatus for mapping analysis by a limited number of electrodes  
US8781566B2 (en)  System and methods for slidingscale cardiac event detection  
JP4602354B2 (en)  Cardiac monitoring  
US6169919B1 (en)  System and method for quantifying alternation in an electrocardiogram signal  
Lu et al.  Can photoplethysmography variability serve as an alternative approach to obtain heart rate variability information?  
US5033472A (en)  Method of and apparatus for analyzing propagation of arterial pulse waves through the circulatory system  
EP1507474B1 (en)  Procedure for deriving reliable information on respiratory activity from heart period measurement  
US6647287B1 (en)  Dynamic cardiovascular monitor  
US6829501B2 (en)  Patient monitor and method with noninvasive cardiac output monitoring  
US6350242B1 (en)  Respiration monitoring system based on sensed physiological parameters  
US20030117296A1 (en)  Method and apparatus for multiple patient parameter variability analysis and display  
EP1446192B1 (en)  Cardiac pacing parameter optimization and monitoring of cardiac dysfunction  
US5443073A (en)  System and method of impedance cardiography monitoring 
Legal Events
Date  Code  Title  Description 

AS  Assignment 
Owner name: MEDIWAVE STAR TECHNOLOGY INC, NORTH CAROLINA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:STAROBIN, JOSEPH M.;CHERNYAK, YURI B.;REEL/FRAME:012710/0197;SIGNING DATES FROM 20020219 TO 20020222 

FPAY  Fee payment 
Year of fee payment: 4 

REMI  Maintenance fee reminder mailed  
REMI  Maintenance fee reminder mailed  
LAPS  Lapse for failure to pay maintenance fees  
STCH  Information on status: patent discontinuation 
Free format text: PATENT EXPIRED DUE TO NONPAYMENT OF MAINTENANCE FEES UNDER 37 CFR 1.362 

FP  Expired due to failure to pay maintenance fee 
Effective date: 20120727 